# astrojones-mcp-retrieval - Doramagic AI Context Pack

> Positioning: a pre-install experience and judgment asset. It helps the host AI get off to a good start, but it does not mean the project has already been installed, run, or validated.

## Sufficiency Principle

- **Sufficiency over compression**: The AI Context Pack should be sufficient for the host AI to understand the project's value, capability boundaries, entrypoints, risks, and evidence sources before starting work; it may be layered, but it does not aim for the shortest possible summary.
- **Compression policy**: Compress only noise and duplication, never context that affects judgment or the quality of the work.

## How the Host AI Should Use This

You are reading the AI Context Pack that Doramagic compiled for astrojones-mcp-retrieval. Treat it as pre-work context: help the user understand who it fits, what it can do, how to start, what must be verified after install, and where the risks are. Do not claim that you have already installed, run, or executed the target project.

## Claim Consumption Rules

- **Fact source**: Repo Evidence + Claim/Evidence Graph; the Human Wiki only supplies salience, terminology, and narrative structure.
- **Minimum status for a fact**: `supported`
- `supported`: May be used as a project fact, but the answer must cite the claim_id and evidence path.
- `weak`: Usable only as a low-confidence lead; the user must be asked to keep verifying.
- `inferred`: Usable only for risk notes or open questions; must not be packaged as a project fact.
- `unverified`: Must not be used as fact; state clearly that evidence is insufficient.
- `contradicted`: Must show the conflicting sources and must not force a single version on the user's behalf.

## Who It Fits Best

- **Developers already using host AIs such as Claude/Codex/Cursor/Gemini**: The README or plugin config mentions multiple host AIs. Evidence: `README.md` Claim: `clm_0004` supported 0.86
- **Users who want to bring professional workflows into a host AI**: The repo contains Skill documents. Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-graph-tune/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-mem-ingest-wisely/SKILL.md`, `skills/bugfix/SKILL.md` et al. Claim: `clm_0005` supported 0.86

## What It Can Do

- **AI Skill / Agent Instruction Asset Library** (Previewable before install): The project contains Skill or Agent instruction files that a host AI can read, useful for bringing professional workflows into hosts like Claude, Codex, or Cursor. Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-graph-tune/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-mem-ingest-wisely/SKILL.md`, `skills/bugfix/SKILL.md` et al. Claim: `clm_0001` supported 0.86
- **Multi-Host Install and Distribution** (Verify after install): The project contains plugin or marketplace configuration, indicating it targets install and distribution across one or more AI hosts. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json` Claim: `clm_0002` supported 0.86
- **Command-Line Startup or Install Flow** (Verify after install): The project documentation contains runnable commands; real use requires running them in a local or host environment. Evidence: `README.md` Claim: `clm_0003` supported 0.86

## How to Start

- `/plugin marketplace add astrojones/claude-plugins` Evidence: `README.md` Claim: `clm_0006` supported 0.86
- `/plugin install astrojones@astrojones` Evidence: `README.md` Claim: `clm_0007` supported 0.86

## Continue-or-Stop Decision Card

- **Current recommendation**: Sandbox-trial the workflow first
- **Why**: This project changes the host AI's development workflow and rules. It is worth trying, but do not install it straight into your primary Claude/Cursor/Codex; verify it first with a temporary host or an isolated directory.

### 30-Second Read

- **What to do now**: Sandbox-trial the workflow first
- **Minimum safe next step**: Feel the workflow constraints with Prompt Preview first; trial in a temporary host only once satisfied
- **Do not trust yet**: Whether this workflow fits your way of working cannot be trusted directly.
- **Continuing will touch**: Host behavior change, Command execution, Host AI configuration

### What You Can Trust Now

- **Target-audience signal: Developers already using host AIs such as Claude/Codex/Cursor/Gemini** (supported): Backed by a supported claim or project evidence, but that still is not the same as real install results. Evidence: `README.md` Claim: `clm_0004` supported 0.86
- **Target-audience signal: Users who want to bring professional workflows into a host AI** (supported): Backed by a supported claim or project evidence, but that still is not the same as real install results. Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-graph-tune/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-mem-ingest-wisely/SKILL.md`, `skills/bugfix/SKILL.md` et al. Claim: `clm_0005` supported 0.86
- **Capability exists: AI Skill / Agent Instruction Asset Library** (supported): You can trust that the project contains signals of this capability; whether it fits your specific task still needs trial or after-install verification. Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-graph-tune/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-mem-ingest-wisely/SKILL.md`, `skills/bugfix/SKILL.md` et al. Claim: `clm_0001` supported 0.86
- **Capability exists: Multi-Host Install and Distribution** (supported): You can trust that the project contains signals of this capability; whether it fits your specific task still needs trial or after-install verification. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json` Claim: `clm_0002` supported 0.86
- **Capability exists: Command-Line Startup or Install Flow** (supported): You can trust that the project contains signals of this capability; whether it fits your specific task still needs trial or after-install verification. Evidence: `README.md` Claim: `clm_0003` supported 0.86
- **There are Quick Start / install-command signals** (supported): You can trust that the docs mention a startup or install entrypoint; do not run it directly in your primary environment because of that. Evidence: `README.md` Claim: `clm_0006` supported 0.86

### What You Cannot Trust Yet

- **Whether this workflow fits your way of working cannot be trusted directly.** (unverified): Workflow-style Skills can strongly constrain AI behavior; they can improve discipline but may also slow your current task pace. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `AGENTS.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md` et al.
- **That it will not conflict with your existing Claude/Cursor/Codex rules cannot be trusted directly.** (inferred): A dev-workflow Skill changes default behaviors like clarifying, planning, testing, and verifying, so it must be tried in a temporary host. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `AGENTS.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md` et al.
- **Real output quality cannot be trusted before install.** (unverified): Prompt Preview can only show how it guides you; it cannot prove result quality in the real project.
- **Host AI version compatibility cannot be trusted before install.** (unverified): Host loading rules and version differences across Claude, Cursor, Codex, Gemini, and others must be verified in a real environment.
- **That it will not pollute your existing host AI's behavior cannot be trusted directly.** (inferred): Skill, plugin, and AGENTS/CLAUDE/GEMINI instructions may change the host AI's default behavior. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `AGENTS.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md` et al.
- **Safe rollback cannot be assumed by default.** (unverified): Unless the project clearly provides uninstall and recovery instructions, verify in an isolated environment first.
- **After a real install, is it compatible with the user's current host AI version?** (unverified): Compatibility can only be verified in the actual host environment. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`
- **Does the project's output quality meet the user's specific task?** (unverified): The pre-install preview can only show flow and boundaries; it cannot replace real evaluation.

### What Continuing Will Touch

- **Host behavior change**: The default development rhythm of clarifying, planning, TDD, verifying, and wrapping up. Why: This kind of Skill draws both its value and its risk from strongly constraining the workflow; confirm first that you are willing to have it change how you work. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `AGENTS.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md` et al.
- **Command execution**: Package managers, network downloads, the local plugin directory, project config, or the user's home directory. Why: Running the very first command can already change your environment; decide whether it is worth running first. Evidence: `README.md`
- **Host AI configuration**: The plugin, Skill, or rule-loading config of hosts like Claude/Codex/Cursor/Gemini/OpenCode. Why: Host configuration changes how the AI works afterward and may conflict with the user's existing rules. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `AGENTS.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md` et al.
- **Local environment or project files**: Install results, plugin caches, project config, or local dependency directories. Why: The write scope and rollback path cannot be proven before install and need isolated verification. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `README.md`
- **Host AI context**: The AI Context Pack, Prompt Preview, Skill routing, risk rules, and project facts. Why: Importing context affects the host AI's later judgment, so avoid packaging unverified items as facts.

### Minimum Safe Next Steps

- **Run Prompt Preview first**: Feel how it changes the AI's development rhythm first, then decide whether to let it into a real host. (applies when: Applies to any project, especially when output quality is unknown.)
- **Trial-install only in an isolated directory or a test account**: Avoid letting install commands pollute your primary host AI, real projects, or home directory. (applies when: When there are signals of command execution, plugin config, or local writes.)
- **Back up your host AI configuration first**: Skill, plugin, and rule files may change the default behavior of Claude/Cursor/Codex. (applies when: When there is a plugin manifest, a Skill, or a host rule entrypoint.)
- **After install, verify just one minimal task**: Verify loading, compatibility, output quality, and rollback first, then decide whether to use it deeply. (applies when: When moving from a trial into a real workflow.)

### Exit Plan

- **Preserve the pre-install state**: Record the original host config and project state so you can later judge whether it is recoverable.
- **Be ready to remove the host plugin / Skill / rule entrypoint**: If behavior is off after the trial install, you can restore the host AI to its pre-trial state.
- **Record the install commands and written paths**: Without clear uninstall instructions, you at least need to know which directories or configs to clean up manually.
- **If there is no rollback path, do not enter your primary environment**: No rollback is a blocker before continuing; do not proceed on trust or luck.

## What Can Only Be Previewed

- Explain who the project fits and what it can do
- Demonstrate a typical conversation flow based on project docs
- Help the user decide whether it is worth installing or researching further

## What Must Be Verified After Install

- Actually installing the Skill, plugin, or CLI
- Running scripts, modifying local files, or accessing external services
- Verifying real output quality, performance, and compatibility

## Boundary & Risk Decision Card

- **Mistaking the pre-install preview for a real run**: The user may overestimate how much configuration, permission, and compatibility verification the project has already done. Mitigation: Clearly separate prompt_preview_can_do from runtime_required. Claim: `clm_0008` inferred 0.45
- **Host AI plugin or Skill rule conflicts**: New rules may change how the user's existing host AI behaves. Mitigation: Inspect the plugin manifest and Skill files before installing, and test in isolation if needed. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json` Claim: `clm_0009` supported 0.86
- **Command execution will modify the local environment**: Install commands may write to the user's home directory, the host plugin directory, or project configuration. Mitigation: Run in an isolated environment or a test account first. Evidence: `README.md` Claim: `clm_0010` supported 0.86
- **To confirm**: After a real install, is it compatible with the user's current host AI version?. Why: Compatibility can only be verified in the actual host environment.
- **To confirm**: Does the project's output quality meet the user's specific task?. Why: The pre-install preview can only show flow and boundaries; it cannot replace real evaluation.
- **To confirm**: Do the install commands require network access, permissions, or global writes?. Why: This affects install risk in both enterprise and personal environments.

## Pre-Work Working Context

### Loading Order

- First read how_to_use.host_ai_instruction to establish the boundaries of this pre-install judgment asset.
- Read claim_graph_summary to confirm facts come from the Claim/Evidence Graph, not the Human Wiki narrative.
- Then read intended_users, capabilities, and quick_start_candidates to judge whether the user is a match.
- When you need to carry out a concrete task, check role_skill_index first, then evidence_index.
- For real install, file modification, network access, performance, or compatibility questions, turn to risk_card and boundaries.runtime_required.

### Task Routes

- **AI Skill / Agent Instruction Asset Library**: Use role_skill_index / evidence_index to help the user pick a usable role, Skill, or workflow first. Boundary: Can be experienced via a pre-install Prompt. Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-graph-tune/SKILL.md`, `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-mem-ingest-wisely/SKILL.md`, `skills/bugfix/SKILL.md` et al. Claim: `clm_0001` supported 0.86
- **Multi-Host Install and Distribution**: State that this is an after-install capability first, then give a pre-install checklist. Boundary: Must be verified after a real install or run. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json` Claim: `clm_0002` supported 0.86
- **Command-Line Startup or Install Flow**: State that this is an after-install capability first, then give a pre-install checklist. Boundary: Must be verified after a real install or run. Evidence: `README.md` Claim: `clm_0003` supported 0.86

### Context Scale

- Total files: 141
- Important-file coverage: 40/141
- Evidence index entries: 80
- Role / Skill entries: 11

### Handling Insufficient Evidence

- **missing_evidence**: State that evidence is insufficient and ask the user for the target file, a README section, or after-install verification records; do not fill in facts.
- **out_of_scope_request**: State that the task is beyond the current AI Context Pack's evidence scope and suggest the user check the Human Manual or verify after a real install.
- **runtime_request**: Provide a pre-install checklist and command sources, but do not run commands for the user or claim they have been run.
- **source_conflict**: Show the conflicting sources side by side, mark them as unverified, and do not force a single version.

## Prompt Recipes

### Fit assessment

- Goal: Judge whether this project fits the user's current task.
- Expected output: A fit conclusion, key reasons, evidence citations, what can be previewed before install, what must be verified after install, and a next-step recommendation.

```text
Based on the AI Context Pack for astrojones-mcp-retrieval, ask me 3 necessary questions first, then judge whether it fits my task. The answer must cover: who it fits, what it can do, what it cannot do, whether it is worth installing, and where the evidence comes from. Every project fact must cite evidence_refs, source_paths, or a claim_id.
```

### Pre-install experience

- Goal: Let the user feel the core workflow before installing, while avoiding packaging the preview as real capability or a marketing promise.
- Expected output: An experience script with boundary labels, an after-install verification checklist, and a cautious recommendation; with no real-run promises or strong marketing language.

```text
Treat astrojones-mcp-retrieval as a pre-install experience asset, not an already-installed tool or a real runtime environment.

Output exactly four parts:
1. Ask me 3 necessary questions first.
2. Give an "experience script": use the three labels [Previewable before install], [Must verify after install], and [Insufficient evidence] to show how it might guide the workflow.
3. Give an after-install verification checklist: list which capabilities can only be confirmed after a real install, real host loading, and a real project run.
4. Give a cautious recommendation: only "worth researching/trialing further", "add information before deciding", or "not recommended to continue"; do not endorse the project.

Hard boundaries:
- Do not claim you have installed, run, executed tests, modified files, or produced real results.
- Do not write promise-like phrasing such as "auto-adapts", "guarantees passing", "perfect fit", or "strongly recommend installing".
- If you describe how it works after install, you must use a conditional such as "if installed successfully and the host loads the Skill correctly, it might...".
- The experience script may only be written as "example lines / hypothetical flow": use "might ask / might suggest / might show", not "has written, has generated, has passed, is running, is generating".
- Prompt Preview does not hand out install commands; if the user is ready to trial, only prompt them to read Quick Start and the Risk Card first and to verify in an isolated environment.
- Every project fact must come from a supported claim, evidence_refs, or source_paths; inferred/unverified items can only be risks or open questions.

```

### Role / Skill selection

- Goal: Pick the best-matching asset from the project's roles or Skills.
- Expected output: A list of candidate roles or Skills, each with an applicable scenario, evidence paths, risk boundary, and whether after-install verification is needed.

```text
Read role_skill_index and recommend 3-5 of the most relevant roles or Skills for my target task. For each recommendation, state the applicable scenario, likely output, risk boundary, and evidence_refs.
```

### Risk pre-check

- Goal: Identify environment, permission, rule-conflict, and quality risks before installing or adopting.
- Expected output: A checklist of environment, permission, dependency, license, host-conflict, quality risk, and unknown items.

```text
Based on risk_card, boundaries, and quick_start_candidates, give me a pre-install risk pre-check list. Do not run commands for me; only explain what I should check, why, and what impact a failure would have.
```

### Host AI kickoff instruction

- Goal: Turn the project context into a host AI instruction for the start of a conversation.
- Expected output: A pre-work instruction with clear boundaries and clear evidence citations, suitable to copy to a host AI.

```text
Based on the AI Context Pack for astrojones-mcp-retrieval, generate a pre-work instruction I can paste to my host AI. This instruction must obey not_runtime=true and must not claim the project has been installed, run, or produced real results.
```

## Role / Skill Index

- Indexed 11 role / Skill / project-doc entries.

- **astrojones-cognee-doctor** (skill): Use when durable memory misbehaves in a repository that has the repo-agent-harness — mem search returns nothing or errors, mem remember seems to vanish, recall is missing at session start, or you suspect double-written memory. Runs the checkable diagnosis mem doctor and walks the known non-obvious cognee deployment gotchas. Activation hint: When the user's task is highly relevant to the workflow described by “astrojones-cognee-doctor”, use it for a pre-install experience first, then decide whether to install. Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md`
- **astrojones-graph-tune** (skill): Use when the cognee memory graph's entity types have gone noisy in a repository that has the repo-agent-harness — mem search returns diluted or scattered results, the same concept appears under many ad-hoc type names, or after a large uncurated ingest. Designs a fixed type vocabulary as NamedIndividuals, pins it with mem ontology, and verifies the collapse with a canary re-ingest. Activation hint: When the user's task is highly relevant to the workflow described by “astrojones-graph-tune”, use it for a pre-install experience first, then decide whether to install. Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-graph-tune/SKILL.md`
- **astrojones-mem-ingest-wisely** (skill): Use when loading a batch of documents, notes, or history into the cognee memory graph in a repository that has the repo-agent-harness — docs folders, ADRs, session digests, incident writeups. Curates before ingesting, cost-checks with a dry run, tags with node sets, and proves retrieval with three canary queries before declaring success. Activation hint: When the user's task is highly relevant to the workflow described by “astrojones-mem-ingest-wisely”, use it for a pre-install experience first, then decide whether to install. Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-mem-ingest-wisely/SKILL.md`
- **bugfix** (skill): Use when fixing a bug, diagnosing a failure, or chasing a stack trace in a repository that has the repo-agent-harness. Guides a safe, minimal-surface fix using Serena for navigation and the harness tools for context and verification. Activation hint: When the user's task is highly relevant to the workflow described by “bugfix”, use it for a pre-install experience first, then decide whether to install. Evidence: `skills/bugfix/SKILL.md`
- **commit** (skill): Creates semantic commits from working tree changes, grouping related changes chronologically and semantically. Plans commit groups with the main agent, then fans out to a swarm of Haiku subagents that draft commit messages in parallel before the main agent commits sequentially. Use when the user asks to clean up git status, create semantic commits, or organize changes into logical commits. Activation hint: When the user's task is highly relevant to the workflow described by “commit”, use it for a pre-install experience first, then decide whether to install. Evidence: `skills/commit/SKILL.md`
- **feature** (skill): Use when adding a new feature or capability to a repository that has the repo-agent-harness. Guides a smallest-vertical-slice implementation modeled on existing patterns, with tests and targeted verification. Activation hint: When the user's task is highly relevant to the workflow described by “feature”, use it for a pre-install experience first, then decide whether to install. Evidence: `skills/feature/SKILL.md`
- **implement** (skill): Use when taking a task from spec to done in a repository that has the repo-agent-harness — a task file, an issue, or a clear inline description. Runs the repo's end-to-end pipeline — spec gate → plan → implement via parallel TDD agents strict RED/GREEN/REFACTOR → verify → ship — wired to the harness tools and bundled subagents. Activation hint: When the user's task is highly relevant to the workflow described by “implement”, use it for a pre-install experience first, then decide whether to install. Evidence: `skills/implement/SKILL.md`
- **onboard** (skill): Use once per project a project may span multiple repos that share one dataset that has the repo-agent-harness to build its durable project memory in the cognee graph — the one-time onboarding that derives a type ontology from the repo's own symbol map and structure, curates and cost-gates the initial memories tech stack, commands, conventions, structure , ingests them into the project dataset under user confirmation… Activation hint: When the user's task is highly relevant to the workflow described by “onboard”, use it for a pre-install experience first, then decide whether to install. Evidence: `skills/onboard/SKILL.md`
- **plan** (skill): Use when planning a task in a repository that has the repo-agent-harness — turning a task file, an issue, or an inline description into an approved implementation plan before any code is written. Runs the harness-native plan flow the explorer and read-only architect subagents, Serena-powered in place of the built-in Explore / Plan agents, and owns the plan-mode gate end to end. Invoked as /astrojones:plan . Activation hint: When the user's task is highly relevant to the workflow described by “plan”, use it for a pre-install experience first, then decide whether to install. Evidence: `skills/plan/SKILL.md`
- **refactor** (skill): Use when restructuring or cleaning up code without changing behavior in a repository that has the repo-agent-harness. Enforces a behavior-preserving, scope-limited refactor with impact analysis and verification. Activation hint: When the user's task is highly relevant to the workflow described by “refactor”, use it for a pre-install experience first, then decide whether to install. Evidence: `skills/refactor/SKILL.md`
- **test** (skill): Use when writing, repairing, or running tests in a repository that has the repo-agent-harness. Guides narrow, targeted testing and disciplined failure triage. Activation hint: When the user's task is highly relevant to the workflow described by “test”, use it for a pre-install experience first, then decide whether to install. Evidence: `skills/test/SKILL.md`

## Evidence Index

- Indexed 80 evidence entries.

- **Marketplace** (structured_config): { "name": "astrojones-dev", "owner": { "name": "astrojones", "url": "https://github.com/astrojones" }, "metadata": { "description": "Self-hosting marketplace for developing the astrojones plugin from this repo's working-copy HEAD. Register it as a local marketplace and enable astrojones@astrojones-dev to run the local source instead of the released version." }, "plugins": { "name": "astrojones", "source": "./", "category": "developer-tools", "keywords": "harness", "mcp", "serena", "coding-agent", "safety", "code-navigation" , "description": "The repo agent harness as a Claude Code plugin, loaded from local HEAD: a bundled, auto-connecting MCP server repo tools + proxied serena code navigati… Evidence: `.claude-plugin/marketplace.json`
- **Plugin** (structured_config): { "name": "astrojones", "version": "3.23.0", "description": "The repo agent harness as a Claude Code plugin: a bundled, auto-connecting MCP server deterministic repo tools + proxied serena code navigation , safety hooks, and generic coding-workflow skills and subagents bugfix/feature/refactor/test/implement; explorer/implementer/reviewer/test-runner/fullstack-architect . Harnesses any repo automatically on connect AGENTS.md is opt-out ; /harness-init is the explicit fallback for non-MCP clients.", "author": { "name": "astrojones", "url": "https://github.com/astrojones" }, "homepage": "https://github.com/astrojones/astrojones", "keywords": "harness", "mcp", "serena", "coding-agent", "safety"… Evidence: `.claude-plugin/plugin.json`
- **Agent guide — astrojones** (documentation): Working in this repo repo-agent-harness Evidence: `AGENTS.md`
- **astrojones** (documentation): The repo agent harness as a Claude Code plugin. Install it once and every git repo you open gets safe, deterministic, repo-aware tooling and symbol-level code navigation — instead of a coding agent improvising with raw shell, and with no harness files written into the repo . Evidence: `README.md`
- **astrojones — opencode plugin** (documentation): The opencode half of the dual-target astrojones harness plugin. Mirrors what the Claude Code half does, except: opencode does not auto-load .claude/ , so this plugin materializes the per-assistant surfaces skills, commands, agents into the locations opencode reads from. Evidence: `opencode/README.md`
- **repo-agent-harness MCP server + CLI** (documentation): repo-agent-harness MCP server + CLI Evidence: `servers/harness-mcp/README.md`
- **Agent guide — REPO NAME** (documentation): Working in this repo repo-agent-harness Evidence: `servers/harness-mcp/repo_agent_harness/templates/AGENTS.md`
- **Package** (package_manifest): { "name": "astrojones-opencode", "version": "0.7.0", "type": "module", "description": "opencode half of the astrojones plugin: materializes workflow skills/commands/agents from the harness MCP server SSOT and wires the policy hook.", "license": "MIT", "main": "plugin/astrojones.ts", "exports": { ".": { "import": "./plugin/astrojones.ts", "types": "./plugin/astrojones.ts" } }, "files": "plugin", "opencode.json", "README.md" , "scripts": { "test": "vitest run" }, "peerDependencies": { "@opencode-ai/plugin": " =1.15.0" }, "devDependencies": { "@opencode-ai/plugin": "1.15.13", "@types/node": "^20.0.0", "typescript": "^5.4.0", "vitest": "^2.1.9" } } Evidence: `opencode/package.json`
- **cognee-doctor — diagnose the durable-memory stack** (skill_instruction): cognee-doctor — diagnose the durable-memory stack Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-cognee-doctor/SKILL.md`
- **graph-tune — pin the memory graph's type vocabulary** (skill_instruction): graph-tune — pin the memory graph's type vocabulary Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-graph-tune/SKILL.md`
- **mem-ingest-wisely — curated, cost-gated bulk memory loads** (skill_instruction): mem-ingest-wisely — curated, cost-gated bulk memory loads Evidence: `servers/harness-mcp/repo_agent_harness/.agents/skills/astrojones-mem-ingest-wisely/SKILL.md`
- **Bugfix workflow** (skill_instruction): 1. Reproduce / locate the failure — the error message, failing test, or stack trace. 2. Find the relevant code - Unfamiliar or multi-file territory? Dispatch the explorer subagent to map the relevant symbols and hand back a focused reading list — it absorbs the file noise instead of flooding this session. - A known single symbol? Go direct: Serena find symbol / find referencing symbols , with repo context relevant files for a heuristic shortlist. 3. Read only what you need with repo read range — never dump whole files. 4. Assess blast radius — if the fix touches a shared symbol, run repo impact file first. 5. Fix the root cause with the smallest possible edit. No unrelated cleanup. 6. Verif… Evidence: `skills/bugfix/SKILL.md`
- **Commit Semantic** (skill_instruction): Create semantic, chronological commits from working tree changes. Plans commit groups first, then fans out to Haiku subagents for parallel message-drafting, and commits sequentially in the correct order. Evidence: `skills/commit/SKILL.md`
- **Feature workflow** (skill_instruction): 1. Orient — repo context overview for languages, entrypoints, and important paths. 2. Find a comparable feature already in the codebase and mirror its structure. In an unfamiliar area, dispatch the explorer subagent to locate it and return a reading list; in a familiar one, go direct with Serena + repo context relevant files . 3. Identify the files involved: API/interface, domain/logic, and tests. 4. Implement the smallest vertical slice that delivers value end-to-end. 5. Add or update tests alongside the code. 6. Verify with repo verify changed ; review repo diff current . Evidence: `skills/feature/SKILL.md`
- **implement — Spec-Driven TDD Pipeline** (skill_instruction): implement — Spec-Driven TDD Pipeline Evidence: `skills/implement/SKILL.md`
- **onboard — one-time durable-memory onboarding for a harness repo** (skill_instruction): onboard — one-time durable-memory onboarding for a harness repo Evidence: `skills/onboard/SKILL.md`
- **plan — harness-native plan mode** (skill_instruction): Produces an approved implementation plan using the harness's read-only subagents instead of the built-in Explore / Plan agents. The agents navigate by symbol Serena rather than reading whole files, and they are dispatched workers — they return findings/design and never touch the plan-mode gate. This skill is the only caller of EnterPlanMode / ExitPlanMode , and the only writer of the plan file. Evidence: `skills/plan/SKILL.md`
- **Refactor workflow** (skill_instruction): 1. State the target — the behavior-preserving change you intend rename, extract, move, simplify . 2. Check impact — run repo impact file on the targets; note dependents and test coverage. If the blast radius is unfamiliar, dispatch the explorer subagent to map the dependents by symbol before you touch them. 3. Make the mechanical change only. Resist unrelated cleanup "while I'm here" is how refactors break . 4. Verify with repo verify changed — behavior must be unchanged. 5. Review repo diff current specifically for scope creep; revert anything outside the stated target. Evidence: `skills/refactor/SKILL.md`
- **License** (source_file): Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files the "Software" , to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: Evidence: `LICENSE`
- **Test workflow** (skill_instruction): 1. Identify the likely test target for the change the harness maps changed source files to tests . 2. Run the narrowest test via repo verify changed or agent/tools/test-changed — not the full suite. 3. On failure, classify before fixing: - test bug wrong assertion/setup , - setup/environment bug, - genuine product bug. 4. Fix at the right layer — don't paper over a product bug by weakening the test. 5. Never update snapshots without a concrete reason. Evidence: `skills/test/SKILL.md`
- **Harness Init** (documentation): Set up the agent harness in the current repository so a coding agent has safe, deterministic, repo-aware tooling. The heavy lifting is done by the bundled harness CLI — do not copy files by hand. Evidence: `commands/harness-init.md`
- **/harness-init — one-time bootstrap of the per-repo harness** (documentation): /harness-init — one-time bootstrap of the per-repo harness Evidence: `servers/harness-mcp/repo_agent_harness/prompts/harness-init.md`
- **Settings** (structured_config): { "enabledPlugins": { "astrojones@astrojones": false }, "extraKnownMarketplaces": { "astrojones": { "source": { "source": "github", "repo": "astrojones/claude-plugins" } } } } Evidence: `.claude/settings.json`
- **.Mcp** (structured_config): { "mcpServers": { "repo-agent-harness": { "command": "${CLAUDE PLUGIN ROOT}/run-mcp.sh" } } } Evidence: `.mcp.json`
- **The verify nudge only fires when the perception daemon has no snapshot yet; the** (source_file): Check = tuple str, bool, str ⋮---- def repo hash repo: str - str ⋮---- def state repo dir state: str, repo: str - Path ⋮---- def ingest path: Path, lines: list str - str None ⋮---- session id = None ⋮---- raw lines = path.read text encoding="utf-8", errors="replace" .splitlines ⋮---- raw = raw.strip ⋮---- obj = json.loads raw ⋮---- lines.append raw tolerate non-JSON verbose noise ⋮---- session id = obj.get "session id" ⋮---- def load evidence stream: str - list str ⋮---- lines: list str = session id = ingest Path stream , lines ⋮---- transcripts = sorted Path.home .glob f".claude/projects/ /{session id}.jsonl" ⋮---- def has lines: list str , substrings: str - bool ⋮---- """True when one evi… Evidence: `docker/e2e_verify.py`
- **Run Mcp** (source_file): root="$ dirname "$0" " if -n "${REPO AGENT HARNESS DEV ROOT:-}" && -d "${REPO AGENT HARNESS DEV ROOT}/servers/harness-mcp" ; then root="${REPO AGENT HARNESS DEV ROOT}" fi exec uv run --project "${root}/servers/harness-mcp" repo-agent-harness-mcp "$@" Evidence: `run-mcp.sh`
- **----------------------------------------- coexistence memory-API wrappers P2** (source_file): DEFAULT TIMEOUT S = 30.0 ⋮---- IDEMPOTENT ATTEMPTS = 3 RETRY BACKOFF S = 0.5 ⋮---- NOT CONFIGURED HINT = ⋮---- class CogneeError RuntimeError ⋮---- def init self, message: str, status: int None = None - None ⋮---- class CogneeNotConfiguredError CogneeError ⋮---- def init self - None ⋮---- class CogneeUnavailableError CogneeError ⋮---- """Raised when the server is unreachable or the circuit breaker is open.""" ⋮---- class CogneeAuthError CogneeError ⋮---- """Raised when login fails or a refreshed token is still rejected.""" ⋮---- def remote base url - str None ⋮---- """The remote cognee root URL from COGNEE BASE URL, or None when unset.""" raw = os.environ.get "COGNEE BASE URL" or "" .strip… Evidence: `servers/harness-mcp/repo_agent_harness/cognee_client.py`
- **--------------------------------------------------------------------------- health + auth** (source_file): LOG = logging.getLogger name ⋮---- IN CONTAINER PORT = 8000 DEFAULT CONTAINER = "astrojones-cognee" DEFAULT VOLUME = "astrojones-cognee-data" DEFAULT PORT = 8765 DEFAULT IMAGE = "cognee/cognee:1.4.0" DEFAULT EMAIL = "harness@example.com" ⋮---- DEFAULT BACKEND = "postgres" PG IMAGE = "pgvector/pgvector:pg17" PG CONTAINER = "astrojones-cognee-postgres" PG VOLUME = "astrojones-cognee-pgdata" PG NETWORK = "astrojones-cognee-net" PG USER = "cognee" PG PASSWORD = "cognee" PG DB = "cognee db" PG PORT = 5432 ⋮---- PG MAX CONNECTIONS = "300" PG SHARED BUFFERS = "1GB" ⋮---- SUMMARIZE PROMPT FILE = Path file .parent / "cognee local summarize prompt.txt" SUMMARIZE PROMPT CONTAINER PATH = "/app/cognee/i… Evidence: `servers/harness-mcp/repo_agent_harness/cognee_local.py`
- **Env override for the claude-mem store location tests point this at a fixture DB .** (source_file): LOG = logging.getLogger name ⋮---- POLL SECONDS = 60.0 ⋮---- BATCH SIZE = 20 ⋮---- POLL BUDGET S = 120.0 POLL INTERVAL S = 2.0 ⋮---- STATUS TIMEOUT = "timeout" STATUS ERROR = "error" ⋮---- DATASET UNSAFE = re.compile r" ^a-z0-9 " ⋮---- Env override for the claude-mem store location tests point this at a fixture DB . CLAUDE MEM DB ENV = "CLAUDE MEM DB" ⋮---- def default claude mem db - Path ⋮---- env = os.environ.get CLAUDE MEM DB ENV or "" .strip ⋮---- def dataset for project: str - str ⋮---- class SyncBreaker ⋮---- @property def state self - str ⋮---- def allow self - bool ⋮---- def record success self - None ⋮---- def record failure self - None ⋮---- class CogneeSync ⋮---- ---------------… Evidence: `servers/harness-mcp/repo_agent_harness/cognee_sync.py`
- **Context** (source_file): LANG BY EXT = { ⋮---- PKG MANAGERS = { ⋮---- KNOWN TOOLS = ⋮---- ENTRYPOINT CANDIDATES = ⋮---- BINARY SNIFF = 4096 RG FIELD COUNT = 3 MIN TERM LEN = 2 HIGH CONFIDENCE SCORE = 4 ⋮---- def read manifest root: str - dict ⋮---- p = Path root / "agent" / "manifest.yml" ⋮---- def top level dirs rootp: Path - list str ⋮---- out = c.name + "/" for c in sorted rootp.iterdir if c.is dir and not c.name.startswith "." ⋮---- def entry names d: Path, , dirs: bool, suffix: str = "" - list str ⋮---- """Sorted names of files stem or dirs directly under d ; skips dot/underscore entries.""" ⋮---- out = ⋮---- def harness summary rootp: Path - dict ⋮---- agents md = rootp / "AGENTS.md" guide = None ⋮---- text =… Evidence: `servers/harness-mcp/repo_agent_harness/context.py`
- **No git origin / override: verify the two-segment shape but not the org.** (source_file): PLACEHOLDER RE = re.compile " " + "REPO " + " NAME PKG " + " " ⋮---- WORKFLOW REF SUFFIX = "/.github/.github/workflows/nuk-deploy.yml" SKIP DIRS = {".git", ".venv", "node modules", " pycache ", ".serena", ".pytest cache"} ⋮---- def lines text: str - list str ⋮---- def strip inline comment line: str - str ⋮---- def origin parts root: Path - tuple str None, str None ⋮---- proc = subprocess.run url = proc.stdout.strip ⋮---- slug = url.rstrip "/" .removesuffix ".git" .replace ":", "/" parts = p for p in slug.split "/" if p ⋮---- def repo name root: Path, override: str None - str ⋮---- def origin owner root: Path, override: str None - str None ⋮---- def find manifest root: Path - tuple Path None… Evidence: `servers/harness-mcp/repo_agent_harness/deploy.py`
- **Detect** (source_file): def governing pyproject root: str, files: list str - Path None ⋮---- rootp = Path root start = common ancestor rootp, files cur = start ⋮---- candidate = cur / "pyproject.toml" ⋮---- cur = cur.parent fallback = rootp / "pyproject.toml" ⋮---- def common ancestor rootp: Path, files: list str - Path ⋮---- dirs = rootp / f .parent for f in files ⋮---- common = dirs 0 ⋮---- common = common.parent ⋮---- def pyproject tools pyproject: Path - set str ⋮---- data = tomllib.load fh ⋮---- tool = data.get "tool" ⋮---- def runner tool: str, config dir: Path - list str None ⋮---- def python cmd root: str, files: list str , tool: str, extra: list str , label: str - dict None ⋮---- pyproject = governing pyp… Evidence: `servers/harness-mcp/repo_agent_harness/detect.py`
- **Drift: emit a short hint, not the full diff.** (source_file): FRONTMATTER RE = re.compile r"\A---\n. ?\n---\n", re.DOTALL ⋮---- def strip frontmatter markdown: str - str ⋮---- stripped = FRONTMATTER RE.sub "", markdown, count=1 ⋮---- """Compare a harness prompt body to an on-disk SKILL.md body. Returns a structured result dict: - ok : True if the local body is in sync OR missing. False would mean an actual error condition parse failure, etc. — drift itself is never an error. - name : prompt identifier. - in sync : True iff the bodies match. - severity : "ok" , "warning" drifted , or "info" missing . - message : human-readable summary. - drift : when drifted, a short unified-diff hint truncated to keep the warning message readable . The comparison stri… Evidence: `servers/harness-mcp/repo_agent_harness/drift.py`
- **Git** (source_file): def repo root cwd: str None = None - str None ⋮---- cwd = os.environ.get "CLAUDE PROJECT DIR" res = shell.run "git", "rev-parse", "--show-toplevel" , cwd=cwd, timeout=10 ⋮---- def require root cwd: str None = None - str ⋮---- root = repo root cwd ⋮---- msg = "not inside a git repository" ⋮---- def porcelain root: str - list tuple str, str ⋮---- out = shell.run "git", "status", "--porcelain" , cwd=root, timeout=15 rows = ⋮---- name = name.split " - ", 1 1 ⋮---- def status root: str - dict ⋮---- branch = shell.run "git", "rev-parse", "--abbrev-ref", "HEAD" , cwd=root, timeout=10 ⋮---- last = shell.run "git", "log", "-1", "--pretty=%h %s" , cwd=root, timeout=10 ⋮---- def head root: str - str ⋮… Evidence: `servers/harness-mcp/repo_agent_harness/git.py`
- **----------------------------------------------------------------------------- runners** (source_file): yaml = importlib.import module "yaml" if importlib.util.find spec "yaml" else None ⋮---- CACHE TTL SECONDS = 300 MAX OUTPUT = 4000 ⋮---- class DiagnosticsGateway Protocol ⋮---- def call from thread self, name: str, arguments: dict - object ⋮---- def load config root: str - HealthConfig ⋮---- h = harness home rid = repo id root candidates = ⋮---- data = yaml.safe load path.read text encoding="utf-8" or {} ⋮---- ----------------------------------------------------------------------------- runners ⋮---- def from verify root: str, cfg: HealthCheckConfig - CheckResult ⋮---- """Adapt a verify.run kind result lint/typecheck/test into a CheckResult.""" data = verify.run kind root, cfg.kind skipped… Evidence: `servers/harness-mcp/repo_agent_harness/health.py`
- **Impact** (source_file): HIGH RISK = ⋮---- def looks like test path: str - bool ⋮---- name = Path path .name.lower has marker = "test" in name or "spec" in name ⋮---- def references root: str, stem: str, exclude: str None = None - list str ⋮---- res = shell.run "rg", "-l", "--color=never", r"\b" + re.escape stem + r"\b" , cwd=root, timeout=20 ⋮---- res = shell.run "git", "grep", "-l", "-w", stem , cwd=root, timeout=20 ⋮---- def test files for root: str, stem: str - list str ⋮---- def file impact root: str, path: str - dict ⋮---- stem = Path path .stem refs = references root, stem, exclude=path test targets = sorted {f for f in refs if looks like test f } set test files for root, stem dependents = d for d in refs if… Evidence: `servers/harness-mcp/repo_agent_harness/impact.py`
- **Mem** (source_file): DEFAULT DATASET = "agent sessions" ⋮---- NODE SET PROJECT DOCS = "project docs" NODE SET SESSION DIGEST = "session digest" NODE SET CODE MAP = "code map" TYPE TAG PREFIX = "type:" CONCEPT TAG PREFIX = "concept:" PROJECT TAG PREFIX = "project:" ⋮---- def resolve dataset root: str None - str ⋮---- onboarded = paths.onboarded dataset root ⋮---- CHARS PER TOKEN = 4 CHUNK TOKENS = 1024 USD PER MTOK ENV = "COGNEE INGEST USD PER MTOK" COST LIMIT ENV = "COGNEE INGEST COST LIMIT USD" DEFAULT USD PER MTOK = 5.0 DEFAULT COST LIMIT USD = 1.0 ⋮---- COGNEE PLUGIN DIR = Path "~/.cognee-plugin" SENTINEL RECENT S = 600 ⋮---- def client client: CogneeClient None - CogneeClient ⋮---- def error exc: CogneeErro… Evidence: `servers/harness-mcp/repo_agent_harness/mem.py`
- **Perception** (source_file): LOG = logging.getLogger name ⋮---- RUNNABLE KINDS = {"lint", "typecheck", "test"} ⋮---- COALESCE SECONDS = 0.75 ⋮---- def summary data: dict - str ⋮---- first = next ln for ln in str data.get "output" or "" .splitlines if ln.strip , "failed" ⋮---- def read snapshot root: str - PerceptionSnapshot None ⋮---- """Load the current perception snapshot from disk, or None when absent/unreadable.""" ⋮---- def git state root: str - GitState ⋮---- """Snapshot branch/HEAD/dirty/conflicts for transition detection cheap; safe in a thread .""" st = git.status root ⋮---- def current state root: str - dict ⋮---- """Return the current perception snapshot as a dict for the repo state tool. Falls back to a git… Evidence: `servers/harness-mcp/repo_agent_harness/perception.py`
- **Policies** (source_file): yaml = None ⋮---- DENY REGEXES = ⋮---- REQUIRE CONFIRM = ⋮---- READERS = {"cat", "less", "more", "head", "tail", "bat", "nl", "od", "xxd", "strings"} ⋮---- @dataclass class Limits ⋮---- max files before plan: int = 8 max lines per read: int = 240 max search results: int = 30 max open ranges per task: int = 20 ⋮---- @dataclass class CommandCheck ⋮---- allowed: bool reason: str requires confirmation: bool = False ⋮---- def to dict self - dict ⋮---- def load yaml path: Path None - dict ⋮---- def find config root: str, name: str - Path None ⋮---- h = harness home rid = repo id root candidates = ⋮---- def limits root: str - Limits ⋮---- p = find config root, "context" data = load yaml p if p els… Evidence: `servers/harness-mcp/repo_agent_harness/policies.py`
- **Shown to native Read once onboarded — code reading stays routed through the harness.** (source_file): GATE ENV = "REPO AGENT HARNESS NO SERENA GATE" ⋮---- CODE EXTENSIONS = frozenset ⋮---- UNBOARDED MSG = ⋮---- Shown to native Read once onboarded — code reading stays routed through the harness. BOARDED MSG = ⋮---- def gate disabled - bool ⋮---- """Whether the operator has switched the gate off for this process.""" ⋮---- def is code file path: str Path - bool ⋮---- def is onboarded root: Path - bool ⋮---- mem dir = root / ".serena" / "memories" ⋮---- IMPLEMENTATION CAPABLE EXTENSIONS = frozenset ⋮---- FIND IMPLEMENTATIONS UNSUPPORTED MSG = ⋮---- def implementation unsupported relative path: str - str None ⋮---- suffix = Path relative path .suffix.lower ⋮---- CAPABILITY GATES: dict str, Calla… Evidence: `servers/harness-mcp/repo_agent_harness/serena_gate.py`
- **Server** (source_file): msg = "the 'fastmcp' package is required: uv add fastmcp" ⋮---- LOG = logging.getLogger name ⋮---- serena = gateway.SerenaGateway git.repo root ⋮---- async def cancel task: asyncio.Task None - None ⋮---- def ensure local if enabled - str None ⋮---- async def bring up local sync: cognee sync.CogneeSync - None ⋮---- base = await asyncio.to thread ensure local if enabled ⋮---- rebuilt = cognee client.get client ⋮---- @asynccontextmanager async def lifespan app: FastMCP - AsyncIterator dict ⋮---- = app root = git.repo root ⋮---- perception daemon = perception.Perception root, gateway= serena if root else None ⋮---- def on repo change changed: set str - None ⋮---- repo watcher = watcher.RepoWatc… Evidence: `servers/harness-mcp/repo_agent_harness/server.py`
- **Symbols** (source_file): LOG = logging.getLogger name ⋮---- INDEX NAME = "symbols.json" INDEX VERSION = 3 GIT TIMEOUT S = 10 PARSE MAX BYTES = 1 000 000 ⋮---- LANG BY EXT = { TS KINDS = { KINDS BY LANG: dict str, dict str, str = { ⋮---- class SymbolRecord BaseModel ⋮---- name: str kind: str start line: int = Field ..., description="1-based first line of the symbol body" end line: int = Field ..., description="1-based last line of the symbol body" parent: str None = Field None, description="enclosing symbol name e.g. a method's class " doc: str None = Field None, description="first line of the symbol's docstring, if any" ⋮---- class FileSymbols BaseModel ⋮---- mtime: float symbols: list SymbolRecord = Field default… Evidence: `servers/harness-mcp/repo_agent_harness/symbols.py`
- **Verify** (source_file): def changed by ext root: str, exts: set str - list str ⋮---- out = ⋮---- p = Path root / f ⋮---- def tests for root: str, py files: list str - list str ⋮---- tracked = set git.ls files root tests: set str = set ⋮---- p = Path f ⋮---- stem = p.stem candidates = ⋮---- def skip msg: str - dict ⋮---- def result res: shell.Result, command: str - dict ⋮---- def group by config root: str, files: list str - dict Path, list str ⋮---- groups: dict Path, list str = {} rootp = Path root ⋮---- pp = detect. governing pyproject root, f config dir = pp.parent if pp else rootp ⋮---- def run grouped root: str, files: list str , select, timeout: int - dict None ⋮---- commands: list str = outputs: list str = o… Evidence: `servers/harness-mcp/repo_agent_harness/verify.py`
- **Watcher** (source_file): DEBOUNCE MS = 500 GIT META SUFFIXES = "/.git/HEAD", "/.git/index" ⋮---- class GitAwareFilter watchfiles.DefaultFilter ⋮---- def call self, change: watchfiles.Change, path: str - bool ⋮---- class RepoWatcher ⋮---- def init self, root: str, on invalidate: Callable set str , None - None ⋮---- async def run self - None ⋮---- paths = self. relevant {p for , p in changes} ⋮---- def stop self - None ⋮---- def relevant self, absolute: set str - set str ⋮---- prefix = self.root.rstrip "/" + "/" relative = {p len prefix : for p in absolute if p.startswith prefix } meta = {p for p in relative if p.startswith ".git/" } ⋮---- def not ignored self, paths: set str - set str ⋮---- unknown = sorted p for p… Evidence: `servers/harness-mcp/repo_agent_harness/watcher.py`
- **Changelog** (documentation): rename: the plugin was renamed from raisl to astrojones . The MCP tool prefix is now mcp plugin astrojones repo-agent-harness was mcp plugin raisl repo-agent-harness , the install ref is astrojones@astrojones was raisl@astrojones , and the command namespace is /astrojones: was /raisl: . The GitHub repo moved to astrojones/astrojones was astrojones/raisl . The MCP server name repo-agent-harness and the Python package repo agent harness are unchanged. Evidence: `CHANGELOG.md`
- **Handover — Restarbeiten v2 2026-07-16** (documentation): Handover — Restarbeiten v2 2026-07-16 Evidence: `HANDOVER-2026-07-16-restarbeiten-v2.md`
- **Your job: read the bodies, then design** (documentation): You are architect . You design implementation plans and weigh architectural trade-offs, and you return a plan a staff engineer would approve. You are the harness-native replacement for the built-in Plan agent : where it would read whole files, you navigate by symbol Serena and precise range the harness , absorbing the file noise in your own context and returning only the plan with path:line citations. Evidence: `agents/architect.md`
- **Your one job: map the relevant symbols** (documentation): You are explorer . You locate the code relevant to a task and return a map of the relevant symbols — the blast radius . You navigate by symbol, not by reading whole files, and you keep the caller's context window clean: you absorb the file noise in your own window and return only a cited reading list. You are read-only and never modify code. Evidence: `agents/explorer.md`
- **Tools** (documentation): You are implementer . You own one stream of a larger task: write the tests and the code for the files assigned to you, and nothing outside that set. Evidence: `agents/implementer.md`
- **Reviewer** (documentation): You are reviewer . Review the current change set; report, do not fix. Evidence: `agents/reviewer.md`
- **Test Cognee Client** (source_file): pytestmark = pytest.mark.anyio ⋮---- @pytest.fixture def anyio backend ⋮---- async def test login is lazy and shared ⋮---- fake = FakeCognee datasets= "kolbe" client = CogneeClient fake.client kwargs ⋮---- async def test expired token refreshes once and succeeds ⋮---- async def test not configured fails closed with hint ⋮---- client = CogneeClient url=None, auth=None, key=None ⋮---- async def test idempotent reads retry on transport error ⋮---- async def test writes are never blind retried ⋮---- async def test http error raises with status ⋮---- fake = FakeCognee ⋮---- async def test circuit opens after threshold and recovers via probe ⋮---- now = 0.0 circuit = CogneeCircuit clock=lambda: n… Evidence: `servers/harness-mcp/tests/test_cognee_client.py`
- **Test Cognee Local** (source_file): @pytest.fixture autouse=True def isolated home tmp path, monkeypatch ⋮---- def completed returncode=0, stdout="", stderr="" ⋮---- def test enabled default auto does not autostart monkeypatch ⋮---- def test enabled off never monkeypatch ⋮---- def test enabled armed requires docker monkeypatch ⋮---- def test docker run args recipe ⋮---- args = cognee local.docker run args "me@example.com", "pw123" ⋮---- def test docker run args mounts summarize prompt when vendored ⋮---- mount = f"{cognee local. SUMMARIZE PROMPT FILE}:{cognee local. SUMMARIZE PROMPT CONTAINER PATH}:ro" ⋮---- def test docker run args skips prompt mount when file absent monkeypatch ⋮---- def test container env mirrors remote em… Evidence: `servers/harness-mcp/tests/test_cognee_local.py`
- **Test Cognee Sync** (source_file): pytestmark = pytest.mark.anyio ⋮---- @pytest.fixture def anyio backend ⋮---- def wired fake: FakeCognee, , never trip: bool = False - CogneeClient ⋮---- kwargs = fake.client kwargs ⋮---- def remember count fake: FakeCognee - int ⋮---- def test dataset defaults to repo basename ⋮---- sync = CogneeSync "/home/dev/astrojones" ⋮---- async def test cycle ships then verifies then records ok tmp path ⋮---- db = store tmp path ⋮---- fake = FakeCognee ledger = SyncLedger sync = CogneeSync str tmp path , db=db, project="myrepo", ledger=ledger ⋮---- seq = path for m, path, p in fake.requests if path in {"/api/v1/remember", "/api/v1/datasets/status"} ⋮---- async def test cycle replay dedup skips alread… Evidence: `servers/harness-mcp/tests/test_cognee_sync.py`
- **Test Context** (source_file): def test overview repo ⋮---- ov = context.overview str repo ⋮---- def test detect languages orders by count repo ⋮---- langs = context.detect languages str repo ⋮---- def test serena languages maps and dedupes repo ⋮---- keys = context.serena languages str repo ⋮---- def test overview configured tools present repo ⋮---- def test overview configured tools reads pyproject repo ⋮---- def test overview harness absent repo ⋮---- h = context.overview str repo "harness" ⋮---- def test overview harness present repo ⋮---- def test read range ok repo ⋮---- out = context.read range str repo , "src/payment.py", 1, 2 ⋮---- def test read range refuses secret repo ⋮---- out = context.read range str repo ,… Evidence: `servers/harness-mcp/tests/test_context.py`
- **Test Deploy** (source_file): DEFAULT OWNER = "astrojones" DEFAULT DOMAIN = "astrojones.de" ⋮---- def set origin repo: Path, owner: str - None ⋮---- def deploy files repo: Path, owner: str = DEFAULT OWNER, domain: str = DEFAULT DOMAIN - None ⋮---- name = repo.name ⋮---- wf = repo / ".github" / "workflows" ⋮---- @pytest.fixture def deployable repo repo: Path - Path ⋮---- def test deploy validate clean repo is ok deployable repo, monkeypatch ⋮---- result = server.repo deploy validate ⋮---- codes = {f "code" for f in result "findings" } ⋮---- def test deploy validate non astrojones org monkeypatch, repo ⋮---- image ok = f for f in result "findings" if f "code" == "image" and f "level" == "ok" ⋮---- workflow ok = f for f in… Evidence: `servers/harness-mcp/tests/test_deploy.py`
- **Test Detect** (source_file): def write p: Path, text: str - None ⋮---- def test pyproject tools reads tool table tmp path: Path ⋮---- pp = tmp path / "pyproject.toml" ⋮---- def test pyproject tools empty when no tool table tmp path: Path ⋮---- def test pyproject tools missing file tmp path: Path ⋮---- def test governing pyproject walks up to nearest tmp path: Path ⋮---- nested = tmp path / "servers" / "pkg" ⋮---- got = detect. governing pyproject str tmp path , "servers/pkg/src/mod.py" ⋮---- def test governing pyproject falls back to root tmp path: Path ⋮---- got = detect. governing pyproject str tmp path , "src/mod.py" ⋮---- def test governing pyproject none when absent tmp path: Path ⋮---- def test python typechecker… Evidence: `servers/harness-mcp/tests/test_detect.py`
- **Test Drift** (source_file): def test compare bodies match returns ok ⋮---- result = drift.compare bodies ⋮---- def test compare bodies drift returns warning ⋮---- def test compare bodies handles missing local ⋮---- def test compare bodies ignores yaml frontmatter ⋮---- def test compare bodies ignores trailing whitespace ⋮---- def test check repo drift all in sync tmp path, monkeypatch ⋮---- entry = prompts registry.get name ⋮---- skill dir = tmp path / "skills" / name ⋮---- result = drift.check repo drift tmp path ⋮---- def test check repo drift detects drifted skill tmp path ⋮---- entry = prompts registry.get "bugfix" ⋮---- skill dir = tmp path / "skills" / "bugfix" ⋮---- def test check repo drift detects missing ski… Evidence: `servers/harness-mcp/tests/test_drift.py`
- **Test Git** (source_file): def test repo root repo ⋮---- def test status clean repo ⋮---- st = git.status str repo ⋮---- def test status dirty repo ⋮---- def test diff current redacts repo ⋮---- d = git.diff current str repo ⋮---- def test repo root uses claude project dir monkeypatch, repo ⋮---- result = git.repo root ⋮---- def test repo root ignores claude project dir when cwd explicit monkeypatch, repo ⋮---- result = git.repo root cwd=str repo ⋮---- def test repo root falls back to process cwd when env unset monkeypatch, repo ⋮---- original = pathlib.Path.cwd ⋮---- def test repo root resolves linked worktree repo, tmp path ⋮---- wt = tmp path / "linked-wt" ⋮---- result = git.repo root cwd=str wt ⋮---- def test rep… Evidence: `servers/harness-mcp/tests/test_git.py`
- **Test Health** (source_file): CHEAP CONFIG = """\ ⋮---- def write config repo: Path, body: str, harness home: Path - None ⋮---- d = harness home / "repos" / repo id str repo ⋮---- def test defaults when no config repo ⋮---- cfg = health.load config str repo ids = {c.id: c for c in cfg.checks} ⋮---- def test invalid config falls back to defaults repo, isolated harness home ⋮---- snap = health.run str repo , only="worktree" ⋮---- def test config selects checks repo, isolated harness home ⋮---- snap = health.run str repo ⋮---- def test git check clean repo repo ⋮---- def test command check pass and fail repo, isolated harness home ⋮---- by id = {c.id: c for c in snap.checks} ⋮---- def test command check policy denied repo,… Evidence: `servers/harness-mcp/tests/test_health.py`
- **Test Impact** (source_file): def test impact high risk repo ⋮---- out = impact.file impact str repo , "src/payment.py" ⋮---- def test impact low risk unreferenced repo ⋮---- out = impact.file impact str repo , "lonely.py" Evidence: `servers/harness-mcp/tests/test_impact.py`
- The remaining 20 evidence entries are in `AI_CONTEXT_PACK.json` or `EVIDENCE_INDEX.json`.

## Rules the Host AI Must Follow

- **Treat this asset as pre-work context, not a runtime environment.**: The AI Context Pack contains only an evidence-backed understanding of the project, not the project's executable state. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `AGENTS.md`
- **When answering the user, distinguish what can be previewed from what can only be verified after install.**: The consumer value of the pre-install experience comes from reducing bad installs and misjudgments, not from pretending to be a real run. Evidence: `.claude-plugin/marketplace.json`, `.claude-plugin/plugin.json`, `AGENTS.md`

## Questions the User Should Answer First

- Which host AI or local environment do you plan to use it in?
- Do you just want to experience the workflow first, or are you ready to actually install?
- What matters most to you: install cost, output quality, or conflicts with your existing rules?

## Acceptance Checks

- Every capability claim can be traced back to a file path in evidence_refs.
- AI_CONTEXT_PACK.md does not package previews as a real run.
- The user can understand who it fits, what it can do, how to start, and the risk boundaries within 3 minutes.

---

## Doramagic Context Augmentation

The following sections strengthen the repository context for a host AI. Human Manual data is a reading route, and pitfall notes become operating constraints.

## Human Manual Outline

Usage rule: this is only a reading route and salience signal, not factual authority. Concrete claims must still return to repo evidence or Claim Graph.

Host AI hard rules:
- Do not treat page titles, section order, summaries, or importance values as factual project evidence.
- When explaining the Human Manual outline, state that it is only a reading route or salience signal.
- Capability, installation, compatibility, runtime state, and risk claims must cite repo evidence, source paths, or Claim Graph.

- **Plugin Overview & System Architecture**: importance `high`
  - source_paths: .claude-plugin/marketplace.json, .claude-plugin/plugin.json, .claude/settings.json, .mcp.json, run-mcp.sh
- **Core Harness Tools, Serena Proxy & Workflow Skills**: importance `high`
  - source_paths: servers/harness-mcp/repo_agent_harness/server.py, servers/harness-mcp/repo_agent_harness/context.py, servers/harness-mcp/repo_agent_harness/impact.py, servers/harness-mcp/repo_agent_harness/git.py, servers/harness-mcp/repo_agent_harness/symbols.py
- **Memory & Perception Layer (cognee ↔ claude-mem)**: importance `high`
  - source_paths: servers/harness-mcp/repo_agent_harness/perception.py, servers/harness-mcp/repo_agent_harness/watcher.py, servers/harness-mcp/repo_agent_harness/mem.py, servers/harness-mcp/repo_agent_harness/cognee_client.py, servers/harness-mcp/repo_agent_harness/cognee_local.py
- **Deployment, Drift Detection & Operations**: importance `high`
  - source_paths: servers/harness-mcp/repo_agent_harness/deploy.py, servers/harness-mcp/repo_agent_harness/drift.py, servers/harness-mcp/repo_agent_harness/verify.py, servers/harness-mcp/repo_agent_harness/health.py, servers/harness-mcp/repo_agent_harness/detect.py

## Repo Inspection Evidence

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `dd3f378d814931b97f4793a70a2ef578e03be23c`
- inspected_files: `README.md`

Host AI hard rules:
- Without repo_clone_verified=true, do not claim that the source code has been read.
- Without repo_inspection_verified=true, do not write README, docs, or package-file conclusions as facts.
- Without quick_start_verified=true, do not claim that the Quick Start path has run successfully.

## Doramagic Pitfall Constraints

These rules come from Doramagic discovery, validation, or compilation findings. The host AI must treat them as operating constraints, not background notes.

### Constraint 1: Capability evidence risk requires verification

- Trigger: README/documentation is current enough for a first validation pass.
- Host AI rule: Reproduce the official install and quickstart path in an isolated environment.
- Why it matters: May increase setup, validation, or first-run risk for the user.
- Evidence: capability.assumptions | https://github.com/astrojones/astrojones
- Hard boundary: Do not present this pitfall as solved, verified, or ignorable unless later evidence explicitly closes it.

### Constraint 2: Security or permission risk requires verification

- Trigger: no_demo
- Host AI rule: Reproduce the official install and quickstart path in an isolated environment.
- Why it matters: May increase setup, validation, or first-run risk for the user.
- Evidence: downstream_validation.risk_items | https://github.com/astrojones/astrojones
- Hard boundary: Do not present this pitfall as solved, verified, or ignorable unless later evidence explicitly closes it.

### Constraint 3: Security or permission risk requires verification

- Trigger: no_demo
- Host AI rule: Reproduce the official install and quickstart path in an isolated environment.
- Why it matters: May increase setup, validation, or first-run risk for the user.
- Evidence: risks.scoring_risks | https://github.com/astrojones/astrojones
- Hard boundary: Do not present this pitfall as solved, verified, or ignorable unless later evidence explicitly closes it.

### Constraint 4: Maintenance risk requires verification

- Trigger: issue_or_pr_quality=unknown。
- Host AI rule: Reproduce the official install and quickstart path in an isolated environment.
- Why it matters: May increase setup, validation, or first-run risk for the user.
- Evidence: evidence.maintainer_signals | https://github.com/astrojones/astrojones
- Hard boundary: Do not present this pitfall as solved, verified, or ignorable unless later evidence explicitly closes it.

### Constraint 5: Maintenance risk requires verification

- Trigger: release_recency=unknown。
- Host AI rule: Reproduce the official install and quickstart path in an isolated environment.
- Why it matters: May increase setup, validation, or first-run risk for the user.
- Evidence: evidence.maintainer_signals | https://github.com/astrojones/astrojones
- Hard boundary: Do not present this pitfall as solved, verified, or ignorable unless later evidence explicitly closes it.