# metaharness - 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 metaharness. 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 - **AI researchers or builders of research-oriented Agents**: The README clearly centers on research, experiment, or paper workflows. Evidence: `README.md` Claim: `clm_0004` supported 0.86 - **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_0005` supported 0.86 - **Users who want to bring professional workflows into a host AI**: The repo contains Skill documents. Evidence: `.claude-plugin/skills/compare-harnesses/SKILL.md`, `.claude-plugin/skills/create-harness/SKILL.md`, `.claude-plugin/skills/diag-harness/SKILL.md`, `.claude-plugin/skills/example-harness/SKILL.md` et al. Claim: `clm_0006` 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: `.claude-plugin/skills/compare-harnesses/SKILL.md`, `.claude-plugin/skills/create-harness/SKILL.md`, `.claude-plugin/skills/diag-harness/SKILL.md`, `.claude-plugin/skills/example-harness/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` unverified 0.25 - **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`, `packages/agent-harness-generator-lib/README.md`, `packages/aws-finops/README.md`, `packages/create-agent-harness/README.md` et al. Claim: `clm_0003` supported 0.86 ## How to Start - `npx metaharness my-bot --template vertical:coding --host claude-code` Evidence: `README.md` Claim: `clm_0007` supported 0.86 - `npx metaharness --wizard` Evidence: `README.md` Claim: `clm_0008` supported 0.86 - `npx metaharness --list` Evidence: `README.md` Claim: `clm_0009` supported 0.86 - `npx metaharness my-bot --template vertical:coding` Evidence: `README.md` Claim: `clm_0007` supported 0.86, `clm_0010` supported 0.86 - `git clone https://github.com/ruvnet/metaharness` Evidence: `README.md` Claim: `clm_0011` supported 0.86 - `npm install @ruvnet/agent-harness-generator` Evidence: `packages/agent-harness-generator-lib/README.md` Claim: `clm_0012` supported 0.86 - `npm install && npm run build && npm test # 34 tests, deterministic` Evidence: `packages/aws-finops/README.md` Claim: `clm_0013` supported 0.86 - `npx metaharness my-bot` Evidence: `packages/create-agent-harness/README.md` Claim: `clm_0007` supported 0.86, `clm_0010` supported 0.86, `clm_0014` supported 0.86, `clm_0018` supported 0.86 - `npx metaharness my-legal-bot \` Evidence: `packages/create-agent-harness/README.md` Claim: `clm_0015` supported 0.86 - `npx metaharness --from-existing ./` Evidence: `packages/create-agent-harness/README.md` Claim: `clm_0016` supported 0.86 ## Continue-or-Stop Decision Card - **Current recommendation**: Trial the research framework first - **Why**: This project targets research workflows; the core risk is source credibility and output quality. Verify the research framework with Prompt Preview first, then trial it in an isolated environment. ### 30-Second Read - **What to do now**: Trial the research framework first - **Minimum safe next step**: Verify the research framework with Prompt Preview first; trial in isolation only once satisfied - **Do not trust yet**: Research conclusions, citations, and experiment results cannot be trusted before install. - **Continuing will touch**: Research judgment, Command execution, Host AI configuration ### What You Can Trust Now - **Target-audience signal: AI researchers or builders of research-oriented Agents** (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: 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_0005` 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: `.claude-plugin/skills/compare-harnesses/SKILL.md`, `.claude-plugin/skills/create-harness/SKILL.md`, `.claude-plugin/skills/diag-harness/SKILL.md`, `.claude-plugin/skills/example-harness/SKILL.md` et al. Claim: `clm_0006` 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: `.claude-plugin/skills/compare-harnesses/SKILL.md`, `.claude-plugin/skills/create-harness/SKILL.md`, `.claude-plugin/skills/diag-harness/SKILL.md`, `.claude-plugin/skills/example-harness/SKILL.md` et al. Claim: `clm_0001` 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`, `packages/agent-harness-generator-lib/README.md`, `packages/aws-finops/README.md`, `packages/create-agent-harness/README.md` et al. 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_0007` supported 0.86 ### What You Cannot Trust Yet - **Research conclusions, citations, and experiment results cannot be trusted before install.** (unverified): A research Skill can organize questions and paths, but it cannot replace real literature search, paper verification, and experiment reproduction. - **Whether it fits your specific research field cannot be trusted directly.** (unverified): The Skill covering many research topics does not mean it is sufficient for your field, source requirements, and credibility standards. - **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`, `.claude-plugin/skills/compare-harnesses/SKILL.md`, `.claude-plugin/skills/create-harness/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 - **Research judgment**: Problem decomposition, source paths, experiment paths, conclusion structure, and credibility judgment. Why: A research Skill can make output look more professional but cannot replace real evidence verification. - **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`, `packages/agent-harness-generator-lib/README.md`, `packages/aws-finops/README.md`, `packages/create-agent-harness/README.md` et al. - **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`, `.claude-plugin/skills/compare-harnesses/SKILL.md`, `.claude-plugin/skills/create-harness/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`, `packages/agent-harness-generator-lib/README.md` et al. - **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**: Verify whether it can correctly frame the research question and evidence boundaries first; do not trust the research output up front. (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. - **Keep a source and conclusion verification checklist**: If citations or experiment paths later prove unreliable, you can return to the evidence-boundary stage and re-check. - **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_0031` 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_0032` inferred 0.45 - **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`, `packages/agent-harness-generator-lib/README.md`, `packages/aws-finops/README.md`, `packages/create-agent-harness/README.md` et al. Claim: `clm_0033` supported 0.86 - **Source document conflict: agent_count**: The project documentation states inconsistent counts; the AI Context Pack must warn the user not to treat any single number as a verified fact. Mitigation: Flag it as unverified in both the Human Manual and the AI Context Pack rather than forcing a single number. Evidence: `docs/adrs/ADR-016-migration-for-ruflo-users.md`, `docs/research/2026-06-21-self-learning-wasm-memory-escalator.md` Claim: `clm_0034` inferred 0.45 - **Source file conflict agent_count**: multiple values `12, 300` found; verify before real use. - **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: `.claude-plugin/skills/compare-harnesses/SKILL.md`, `.claude-plugin/skills/create-harness/SKILL.md`, `.claude-plugin/skills/diag-harness/SKILL.md`, `.claude-plugin/skills/example-harness/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` - **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`, `packages/agent-harness-generator-lib/README.md`, `packages/aws-finops/README.md`, `packages/create-agent-harness/README.md` et al. Claim: `clm_0003` supported 0.86 ### Context Scale - Total files: 3350 - Important-file coverage: 40/3350 - Evidence index entries: 94 - Role / Skill entries: 14 ### 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 metaharness, 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 metaharness 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 metaharness, 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 14 role / Skill / project-doc entries. - **compare-harnesses** (skill): Diff two scaffolded harnesses ADR-031 . Reports manifest meta drift + host list + per-file fingerprint changes added/removed/changed . Exits 0 IDENTICAL, 1 DRIFT, 2 missing manifest. Use --bundle for the ADR-031 schema-1 JSON envelope. Activation hint: When the user's task is highly relevant to the workflow described by “compare-harnesses”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/compare-harnesses/SKILL.md` - **create-harness** (skill): Scaffold your own focused AI agent harness — pick host Claude Code, Codex, pi.dev, Hermes , template, agents, skills, and ship a npm-publishable harness with its own npx CLI. Use when a user asks to "create my own agent harness", "scaffold a harness", "make a custom Claude Code plugin like ruflo", or "build a vertical AI assistant for X". Activation hint: When the user's task is highly relevant to the workflow described by “create-harness”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/create-harness/SKILL.md` - **diag-harness** (skill): Kernel-version skew check ADR-027 . Reports manifest surface + manifest kernel + installed kernel + verdict match/patch-diff/minor-diff/major-diff . Exits 1 on minor/major skew with a copy-pasteable npm install @metaharness/kernel@X.Y.Z next step. Exits 2 if no .harness/manifest.json at path. Activation hint: When the user's task is highly relevant to the workflow described by “diag-harness”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/diag-harness/SKILL.md` - **example-harness** (skill): Scaffold a ready-made AI agent harness in one command from the 19 published @metaharness/ example packages — 9 host integrations Claude Code, Codex, Hermes, pi.dev, OpenClaw, RVM, Copilot, OpenCode, GitHub Actions + 10 vertical pods devops, research, trading, support, legal, coding, education, sales, gaming, repo-maintainer . Activation hint: When the user's task is highly relevant to the workflow described by “example-harness”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/example-harness/SKILL.md` - **harness-secrets** (skill): GCP Secret Manager integration: validate setup, fetch values, or confirm an NPM TOKEN is non-revoked via npm whoami . Used for publish-time token rotation without long-lived keys in CI. Activation hint: When the user's task is highly relevant to the workflow described by “harness-secrets”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/harness-secrets/SKILL.md` - **list-templates** (skill): List the available harness templates and what each one ships with. Use when the user asks "what templates are available", "what verticals does the harness generator support", or "show me what I can scaffold". Activation hint: When the user's task is highly relevant to the workflow described by “list-templates”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/list-templates/SKILL.md` - **oia-manifest** (skill): Emit .harness/oia-manifest.json declaring layer alignment with the OIA v0.1 9-layer reference architecture. Self-describes the harness's MCP wiring, witness signing, audit log, identity posture always 'none' at v0.1 . --check verifies an existing manifest, --dry-run prints without writing, --json emits to stdout. Activation hint: When the user's task is highly relevant to the workflow described by “oia-manifest”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/oia-manifest/SKILL.md` - **publish-harness** (skill): Publish a generated harness to npm — runs the smoke test, signs the witness manifest, and dispatches npm publish --provenance from your tagged release. Activation hint: When the user's task is highly relevant to the workflow described by “publish-harness”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/publish-harness/SKILL.md` - **repo-genome** (skill): 7-section readiness scorecard for a LOCAL repo. Reports repo type + agent topology + MCP risk + test confidence + release readiness + recommended harness plan + scorecard. Exit 0 ready, 1 needs-work, 2 blocked. --json for the 6-field scorecard, --bundle for the ADR-031 schema-1 envelope. Activation hint: When the user's task is highly relevant to the workflow described by “repo-genome”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/repo-genome/SKILL.md` - **score-harness** (skill): 5-dimension scorecard 0-100, grade A/B/C/F for a scaffolded harness. Dimensions: Repo understanding 25% , Agent usefulness 25% , MCP safety 20% , Test coverage 15% , Publish readiness 15% . Emits a 6-field badges block score + mcpRisk + 4 booleans ready for the harness README. Exit 0 A/B, 1 C, 2 F. Activation hint: When the user's task is highly relevant to the workflow described by “score-harness”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/score-harness/SKILL.md` - **threat-model** (skill): MCP threat-model artifact for a scaffolded harness. Reports allowed/denied tools, dangerous permissions count, secrets reachability, network/shell/file-write grants, default-deny posture. Verdict: clean exit 0 / medium exit 1 / high exit 2 . The 'enterprise gold' artifact for PR + compliance review. Activation hint: When the user's task is highly relevant to the workflow described by “threat-model”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/threat-model/SKILL.md` - **upgrade-harness** (skill): Drift detection + apply for a scaffolded harness. Re-renders the template with the same vars, computes added/removed/changed file plan, and applies with Git-style conflict markers or .rej files. Default is dry-run. Activation hint: When the user's task is highly relevant to the workflow described by “upgrade-harness”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/upgrade-harness/SKILL.md` - **validate-harness** (skill): Release-readiness umbrella check for a scaffolded harness — runs doctor, witness verify, hardcoded-path scan, MCP server config, and GCP Secret Manager validation in one shot. Exits non-zero if any sub-check fails. Activation hint: When the user's task is highly relevant to the workflow described by “validate-harness”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/validate-harness/SKILL.md` - **verify-witness** (skill): Verify the Ed25519 witness manifest of a scaffolded harness. Fast yes/no signature check that proves the publisher signed this exact file set — separate from the full release-readiness umbrella in validate-harness. Activation hint: When the user's task is highly relevant to the workflow described by “verify-witness”, use it for a pre-install experience first, then decide whether to install. Evidence: `.claude-plugin/skills/verify-witness/SKILL.md` ## Evidence Index - Indexed 94 evidence entries. - **SaaS Strategy Research** (documentation): Decision-grade research on productizing the MetaHarness/Darwin stack as a commercial service. Evidence: `docs/research/saas/README.md` - **MetaHarness** (documentation): Mint a custom AI agent harness from any repo. Evidence: `README.md` - **@metaharness/ — one-command example harnesses** (documentation): @metaharness/ — one-command example harnesses Evidence: `examples-packages/README.md` - **Examples** (documentation): Real, runnable patterns showing how to use agent-harness-generator . Evidence: `examples/README.md` - **compare-harnesses** (documentation): Codex skill: diff two scaffolded harnesses — the ADR-031 Bundle JSON Pattern surfaced through Codex iter 105 → iter 109 . Evidence: `.codex/skills/compare-harnesses/README.md` - **create-harness Codex skill** (documentation): This skill scaffolds an AI agent harness from inside the OpenAI Codex CLI. Evidence: `.codex/skills/create-harness/README.md` - **diag-harness** (documentation): Codex skill: kernel-version skew check for a scaffolded harness — the ADR-027 diagnostic UX loop. Evidence: `.codex/skills/diag-harness/README.md` - **example-harness Codex skill** (documentation): One-command scaffolding from the 18 published @metaharness/ example packages — the fastest path from a use-case to a working harness, no template/host flags to remember. Evidence: `.codex/skills/example-harness/README.md` - **harness-secrets** (documentation): Codex skill for GCP Secret Manager — check / fetch / validate-token. Evidence: `.codex/skills/harness-secrets/README.md` - **oia-manifest** (documentation): Codex skill: emit/validate the OIA cross-cutting manifest layer ADR-034, iter 121 → iter 122 . Evidence: `.codex/skills/oia-manifest/README.md` - **publish-harness** (documentation): Codex skill that runs the full smoke-test → witness-sign → npm publish pipeline for a generated harness. Evidence: `.codex/skills/publish-harness/README.md` - **repo-genome** (documentation): Codex skill: 7-section readiness scorecard for a local repo — the ADR-031 Bundle JSON Pattern surfaced through Codex iter 110 → 114 . Evidence: `.codex/skills/repo-genome/README.md` - **score-harness** (documentation): Codex skill: 5-dimension harness scorecard with README-ready badges iter 111 → iter 114 . Evidence: `.codex/skills/score-harness/README.md` - **threat-model** (documentation): Codex skill: MCP threat-model artifact for PR / compliance review iter 112 → iter 114 . User-labelled "enterprise gold." Evidence: `.codex/skills/threat-model/README.md` - **upgrade-harness** (documentation): Codex skill: drift detection + apply for a scaffolded harness. Evidence: `.codex/skills/upgrade-harness/README.md` - **validate-harness** (documentation): Codex skill that runs the 5 release-readiness gates from harness validate . Evidence: `.codex/skills/validate-harness/README.md` - **verify-witness** (documentation): Codex skill: Ed25519 witness manifest verification for a scaffolded harness. Evidence: `.codex/skills/verify-witness/README.md` - **Agent Harness Studio — Web UI** (documentation): A 100% client-side Studio for the agent-harness supply chain: turn any GitHub repo or a blank slate into a governed, branded, multi-host AI agent harness — recommend, build, verify, ship. No backend, no install, nothing leaves your browser. Deployable to GitHub Pages; desktop- and mobile-friendly. Evidence: `apps/web-ui/README.md` - **poker-darwin** (documentation): A poker trainer/solver built on three pillars: Evidence: `crates/poker-darwin/README.md` - **MetaHarness × Claude Code workspace + plugin** (documentation): MetaHarness × Claude Code workspace + plugin Evidence: `examples-packages/claude-code/README.md` - **MetaHarness × OpenAI Codex / Codex CLI** (documentation): MetaHarness × OpenAI Codex / Codex CLI Evidence: `examples-packages/codex/README.md` - **MetaHarness: coding vertical** (documentation): ⚠️ Illustrative output. Transcripts and validation/run output shown in this README are representative examples, not captured from a specific run — actual output depends on your environment, models, and inputs. Run the commands to see real results. Evidence: `examples-packages/coding/README.md` - **MetaHarness × VSCode/Copilot mcp.json** (documentation): MetaHarness × VSCode/Copilot mcp.json Evidence: `examples-packages/copilot/README.md` - **MetaHarness: devops vertical** (documentation): ⚠️ Illustrative output. Transcripts and validation/run output shown in this README are representative examples, not captured from a specific run — actual output depends on your environment, models, and inputs. Run the commands to see real results. Evidence: `examples-packages/devops/README.md` - **MetaHarness: education vertical** (documentation): ⚠️ Illustrative output. Transcripts and validation/run output shown in this README are representative examples, not captured from a specific run — actual output depends on your environment, models, and inputs. Run the commands to see real results. Evidence: `examples-packages/education/README.md` - **@metaharness/example-ads** (documentation): Multi-agent scaffold for Google Ads + Meta Marketing API — campaign and spend analysis, read-only by default, mutations gated. Evidence: `examples-packages/example-ads/README.md` - **@metaharness/example-aws** (documentation): AWS infra agent, scaffolded in one command — S3, EC2, Lambda, DynamoDB, STS. Evidence: `examples-packages/example-aws/README.md` - **@metaharness/example-azure** (documentation): MetaHarness scaffold for Microsoft Azure — resource management, Blob storage, and Azure OpenAI via DefaultAzureCredential Evidence: `examples-packages/example-azure/README.md` - **@metaharness/example-bio** (documentation): One command to scaffold a bioinformatics agent harness — gene lookup, PubMed literature retrieval, and genomic sequence fetch, wired to NCBI E-utilities and the Ensembl REST API. Evidence: `examples-packages/example-bio/README.md` - **@metaharness/example-datadog** (documentation): Incident triage from metrics, logs, and monitors — powered by the Datadog API and MetaHarness. Evidence: `examples-packages/example-datadog/README.md` - **@metaharness/example-fhir** (documentation): One-command FHIR R4 agent harness — reads Patient / Observation / Device from a public sandbox EHR, drives multi-agent clinical queries, and verifies every result before reporting done. Evidence: `examples-packages/example-fhir/README.md` - **@metaharness/example-gcp** (documentation): A one-command MetaHarness scaffold wired to Google Cloud Storage, BigQuery, and Vertex AI Gemini — read-only and safe by default. Evidence: `examples-packages/example-gcp/README.md` - **@metaharness/example-github** (documentation): AI-agent harness for GitHub — PR review, issue triage, and release notes, read-only by default. Evidence: `examples-packages/example-github/README.md` - **@metaharness/example-huggingface** (documentation): One command scaffolds a multi-agent Hugging Face harness: model + dataset discovery, serverless inference, and Space exploration — wired to all MetaHarness hosts. Evidence: `examples-packages/example-huggingface/README.md` - **@metaharness/example-iot** (documentation): IoT and robotics telemetry agent with guarded actuation — scaffold in one command. Evidence: `examples-packages/example-iot/README.md` - **@metaharness/example-nasa** (documentation): Space-data agent harness — APOD imagery, EONET Earth events, and satellite pass prediction from TLE Evidence: `examples-packages/example-nasa/README.md` - **MetaHarness x Pinecone — RAG memory agent** (documentation): MetaHarness x Pinecone — RAG memory agent Evidence: `examples-packages/example-pinecone/README.md` - **@metaharness/example-qiskit** (documentation): Scaffold a MetaHarness agent pre-wired to IBM Quantum — build and simulate circuits locally before touching real hardware. Evidence: `examples-packages/example-qiskit/README.md` - **@metaharness/example-slack** (documentation): Channel triage, scoped-token notify, and a /triage slash-command bot — pre-wired to @slack/web-api + @slack/bolt , safe by default, runnable on every metaharness host in one command. Evidence: `examples-packages/example-slack/README.md` - **@metaharness/example-stripe** (documentation): Agent harness scaffold pre-wired to the Stripe billing API — subscriptions, refunds, and webhook handling in TEST MODE by default. Evidence: `examples-packages/example-stripe/README.md` - **@metaharness/example-supabase** (documentation): RLS-aware data agent over Postgres + Auth + Storage — anon vs. service-role keys, pgvector RAG, verification-gated output. Evidence: `examples-packages/example-supabase/README.md` - **@metaharness/example-twilio** (documentation): MetaHarness scaffold for Twilio — SMS, voice, and WhatsApp agents with test-credential sandbox and Messaging Service scoping Evidence: `examples-packages/example-twilio/README.md` - **@metaharness/example-web3** (documentation): A MetaHarness scaffold that wires a multi-agent harness to the Ethereum ecosystem via viem — reading balances, streaming events, and simulating transactions on a testnet, safely by default. Evidence: `examples-packages/example-web3/README.md` - **MetaHarness: gaming vertical** (documentation): A ready-made multi-agent game-design pod scaffolded as a Claude Code harness. One command drops a working claude-code plugin directory on disk with four specialist agents wired up — a concept lead, a mechanics designer, a balance analyst, and a playtest critic — plus the project settings, slash commands, and MCP plumbing needed to actually run them. Aimed at solo designers, game-jam teams, and studios who want a structured pod for ideation and iteration instead of a single generalist prompt. This is a scaffold, not a game engine: it does not ship Unity/Unreal/Godot integrations, no asset pipeline, no live ops — it gives you the design loop, you bring the runtime. Evidence: `examples-packages/gaming/README.md` - **MetaHarness × GitHub Actions** (documentation): A one-command scaffold for a non-interactive agent harness that runs on the GitHub Actions runner — no human at the keyboard. It drops a trigger workflow and a reusable composite action into .github/ , wires least-privilege token permissions, and is ready to run from a webhook manual dispatch, issue comment, push, PR, or schedule . Evidence: `examples-packages/github-actions/README.md` - **MetaHarness × Hermes cli-config harness** (documentation): MetaHarness × Hermes cli-config harness Evidence: `examples-packages/hermes/README.md` - **MetaHarness: legal vertical** (documentation): ⚠️ Illustrative output. Transcripts and validation/run output shown in this README are representative examples, not captured from a specific run — actual output depends on your environment, models, and inputs. Run the commands to see real results. Evidence: `examples-packages/legal/README.md` - **MetaHarness × OpenClaw .openclaw/ config** (documentation): MetaHarness × OpenClaw .openclaw/ config Evidence: `examples-packages/openclaw/README.md` - **MetaHarness × OpenCode .opencode/ config** (documentation): MetaHarness × OpenCode .opencode/ config Evidence: `examples-packages/opencode/README.md` - **MetaHarness × pi.dev AGENTS.md harness** (documentation): MetaHarness × pi.dev AGENTS.md harness Evidence: `examples-packages/pi-dev/README.md` - **MetaHarness: repo-maintainer vertical** (documentation): MetaHarness: repo-maintainer vertical Evidence: `examples-packages/repo-maintainer/README.md` - **MetaHarness: research vertical** (documentation): ⚠️ Illustrative output. Transcripts and validation/run output shown in this README are representative examples, not captured from a specific run — actual output depends on your environment, models, and inputs. Run the commands to see real results. Evidence: `examples-packages/research/README.md` - **MetaHarness × RVM deployment-target partition** (documentation): MetaHarness × RVM deployment-target partition Evidence: `examples-packages/rvm/README.md` - **MetaHarness: sales vertical** (documentation): A ready-made multi-agent sales pipeline pod scaffolded onto Claude Code. Spawns three specialized agents — a qualifier lead scoring , an opener first-touch outreach , and a closer negotiation and follow-through — wired together with shared memory, MCP tooling, and tier-routed model selection. This scaffold gives you a working harness directory you can run, validate, and extend. It does NOT ship a CRM, send live email, or include any contact data — bring your own pipeline source and outbound transport. Evidence: `examples-packages/sales/README.md` - **MetaHarness: support vertical** (documentation): A ready-made customer-support multi-agent template for Claude Code. Scaffolds a four-agent pipeline — triager, KB-searcher, responder, escalator — wired with tier-routed model selection, shared memory, and the MCP servers a support workflow actually needs. It is opinionated about pipeline shape and agent boundaries; it is not a hosted product, not a ticketing system, and not a CRM. You bring the knowledge base, the inbox, and the escalation channel; the scaffold gives you the agents and the glue. Evidence: `examples-packages/support/README.md` - **MetaHarness: trading vertical** (documentation): ⚠️ Illustrative output. Transcripts and validation/run output shown in this README are representative examples, not captured from a specific run — actual output depends on your environment, models, and inputs. Run the commands to see real results. Evidence: `examples-packages/trading/README.md` - **education — runnable demo of vertical:education iter 80** (documentation): education — runnable demo of vertical:education iter 80 Evidence: `examples/education/README.md` - **Federation example** (documentation): Brought up to date in iter 128.2 — published CLI is now metaharness https://www.npmjs.com/package/metaharness . Install via npx metaharness ... . Evidence: `examples/federation/README.md` - **host-tour — scaffold + validate for every supported host** (documentation): host-tour — scaffold + validate for every supported host Evidence: `examples/host-tour/README.md` - **Multi-host example** (documentation): Updated in iter 128.2 to use the published CLI name metaharness https://www.npmjs.com/package/metaharness . Evidence: `examples/multi-host/README.md` - The remaining 34 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: `docs/research/saas/README.md`, `README.md`, `examples-packages/README.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: `docs/research/saas/README.md`, `README.md`, `examples-packages/README.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. - **Introduction & Quick Start**: importance `high` - source_paths: README.md, docs/USERGUIDE.md, docs/USAGE.md, packages/create-agent-harness/src/bin.ts, packages/create-agent-harness/src/subcommands.ts - **System Architecture & Kernel**: importance `high` - source_paths: crates/kernel/src/lib.rs, crates/kernel-wasm/src/lib.rs, crates/kernel-napi/src/lib.rs, packages/kernel-js/src/index.ts, crates/kernel/src/witness.rs - **Templates, Hosts & Plugin Manifests**: importance `high` - source_paths: packages/create-agent-harness/templates/catalog.json, packages/create-agent-harness/templates/vertical_coding/manifest.json, packages/create-agent-harness/templates/minimal/.claude-plugin/plugin.json.tmpl, packages/host-claude-code/src/index.ts, packages/host-codex/src/index.ts - **Self-Evolution, Routing & Cost-Pareto**: importance `high` - source_paths: packages/darwin-mode/src/evolve.ts, packages/darwin-mode/src/scorer.ts, packages/darwin-mode/src/pareto.ts, packages/darwin-mode/src/gepa/loop.ts, packages/router/src/index.ts ## Repo Inspection Evidence - repo_clone_verified: true - repo_inspection_verified: true - repo_commit: `f11fa67454b5f5680af8b8535691f2eade3d6df4` - inspected_files: `README.md`, `package.json`, `docs/ARCHITECTURE.md`, `docs/LOOP_WORKER.md`, `docs/OVERVIEW.md`, `docs/RELEASE.md`, `docs/SOTA_HORIZON.md`, `docs/USAGE.md`, `docs/USERGUIDE.md`, `docs/adrs/ADR-001-goals-and-non-goals.md`, `docs/adrs/ADR-002-kernel-boundary.md`, `docs/adrs/ADR-002a-rust-wasm-napi-publishing-pipeline.md`, `docs/adrs/ADR-003-generator-architecture.md`, `docs/adrs/ADR-004-host-integration-model.md`, `docs/adrs/ADR-005-marketplace-plugin-design.md`, `docs/adrs/ADR-006-memory-and-learning-integration.md`, `docs/adrs/ADR-007-ci-guards.md`, `docs/adrs/ADR-008-drift-detection.md`, `docs/adrs/ADR-009-anti-slop.md`, `docs/adrs/ADR-010-tdd-test-contracts.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/ruvnet/metaharness - Hard boundary: Do not present this pitfall as solved, verified, or ignorable unless later evidence explicitly closes it.