# termyte - Doramagic AI Context Pack > 定位：安装前体验与判断资产。它帮助宿主 AI 有一个好的开始，但不代表已经安装、执行或验证目标项目。 ## 充分原则 - **充分原则，不是压缩原则**：AI Context Pack 应该充分到让宿主 AI 在开工前理解项目价值、能力边界、使用入口、风险和证据来源；它可以分层组织，但不以最短摘要为目标。 - **压缩策略**：只压缩噪声和重复内容，不压缩会影响判断和开工质量的上下文。 ## 给宿主 AI 的使用方式 你正在读取 Doramagic 为 termyte 编译的 AI Context Pack。请把它当作开工前上下文：帮助用户理解适合谁、能做什么、如何开始、哪些必须安装后验证、风险在哪里。不要声称你已经安装、运行或执行了目标项目。 ## Claim 消费规则 - **事实来源**：Repo Evidence + Claim/Evidence Graph；Human Wiki 只提供显著性、术语和叙事结构。 - **事实最低状态**：`supported` - `supported`：可以作为项目事实使用，但回答中必须引用 claim_id 和证据路径。 - `weak`：只能作为低置信度线索，必须要求用户继续核实。 - `inferred`：只能用于风险提示或待确认问题，不能包装成项目事实。 - `unverified`：不得作为事实使用，应明确说证据不足。 - `contradicted`：必须展示冲突来源，不得替用户强行选择一个版本。 ## 它最适合谁 - **正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**：README 或插件配置提到多个宿主 AI。 证据：`README.md` Claim：`clm_0002` supported 0.86 ## 它能做什么 - **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`README.md` Claim：`clm_0001` supported 0.86 ## 怎么开始 - `npm install -g termyte` 证据：`README.md` Claim：`clm_0003` supported 0.86 ## 继续前判断卡 - **当前建议**：需要管理员/安全审批 - **为什么**：继续前可能涉及密钥、账号、外部服务或敏感上下文，建议先经过管理员或安全审批。 ### 30 秒判断 - **现在怎么做**：需要管理员/安全审批 - **最小安全下一步**：先跑 Prompt Preview；若涉及凭证或企业环境，先审批再试装 - **先别相信**：这套流程是否适合你的工作方式不能直接相信。 - **继续会触碰**：宿主行为改变、命令执行、本地环境或项目文件 ### 现在可以相信 - **适合人群线索：正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`README.md` Claim：`clm_0002` supported 0.86 - **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`README.md` Claim：`clm_0001` supported 0.86 - **存在 Quick Start / 安装命令线索**（supported）：可以相信项目文档出现过启动或安装入口；不要因此直接在主力环境运行。 证据：`README.md` Claim：`clm_0003` supported 0.86 ### 现在还不能相信 - **这套流程是否适合你的工作方式不能直接相信。**（unverified）：流程型 Skill 可能强约束 AI 行为；它能提升纪律，也可能拖慢你当前任务节奏。 - **不会和你已有 Claude/Cursor/Codex 规则冲突，不能直接相信。**（inferred）：开发流程 Skill 会改变澄清、计划、测试、验证等默认行为，必须在临时宿主里试。 - **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。 - **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。 - **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。 - **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。 - **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。 - **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。 ### 继续会触碰什么 - **宿主行为改变**：澄清、计划、TDD、验证、收尾等默认开发节奏。 原因：这类 Skill 的价值和风险都来自强约束流程；必须先确认你愿意被它改变工作方式。 - **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`README.md` - **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`README.md` - **环境变量 / API Key**：项目入口文档明确出现 API key、token、secret 或账号凭证配置。 原因：如果真实安装需要凭证，应先使用测试凭证并经过权限/合规判断。 证据：`README.md`, `docs/getting-started.md` - **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。 ### 最小安全下一步 - **先跑 Prompt Preview**：先感受它会怎样改变 AI 的开发节奏，再决定是否让它进入真实宿主。（适用：任何项目都适用，尤其是输出质量未知时。） - **只在隔离目录或测试账号试装**：避免安装命令污染主力宿主 AI、真实项目或用户主目录。（适用：存在命令执行、插件配置或本地写入线索时。） - **不要使用真实生产凭证**：环境变量/API key 一旦进入宿主或工具链，可能产生账号和合规风险。（适用：出现 API、TOKEN、KEY、SECRET 等环境线索时。） - **安装后只验证一个最小任务**：先验证加载、兼容、输出质量和回滚，再决定是否深用。（适用：准备从试用进入真实工作流时。） ### 退出方式 - **保留安装前状态**：记录原始宿主配置和项目状态，后续才能判断是否可恢复。 - **记录安装命令和写入路径**：没有明确卸载说明时，至少要知道哪些目录或配置需要手动清理。 - **准备撤销测试 API key 或 token**：测试凭证泄露或误用时，可以快速止损。 - **如果没有回滚路径，不进入主力环境**：不可回滚是继续前阻断项，不应靠信任或运气继续。 ## 哪些只能预览 - 解释项目适合谁和能做什么 - 基于项目文档演示典型对话流程 - 帮助用户判断是否值得安装或继续研究 ## 哪些必须安装后验证 - 真实安装 Skill、插件或 CLI - 执行脚本、修改本地文件或访问外部服务 - 验证真实输出质量、性能和兼容性 ## 边界与风险判断卡 - **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0004` inferred 0.45 - **命令执行会修改本地环境**：安装命令可能写入用户主目录、宿主插件目录或项目配置。 处理方式：先在隔离环境或测试账号中运行。 证据：`README.md` Claim：`clm_0005` supported 0.86 - **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。 - **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。 - **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。 ## 开工前工作上下文 ### 加载顺序 - 先读取 how_to_use.host_ai_instruction，建立安装前判断资产的边界。 - 读取 claim_graph_summary，确认事实来自 Claim/Evidence Graph，而不是 Human Wiki 叙事。 - 再读取 intended_users、capabilities 和 quick_start_candidates，判断用户是否匹配。 - 需要执行具体任务时，优先查 role_skill_index，再查 evidence_index。 - 遇到真实安装、文件修改、网络访问、性能或兼容性问题时，转入 risk_card 和 boundaries.runtime_required。 ### 任务路由 - **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md` Claim：`clm_0001` supported 0.86 ### 上下文规模 - 文件总数：195 - 重要文件覆盖：40/195 - 证据索引条目：79 - 角色 / Skill 条目：10 ### 证据不足时的处理 - **missing_evidence**：说明证据不足，要求用户提供目标文件、README 段落或安装后验证记录；不要补全事实。 - **out_of_scope_request**：说明该任务超出当前 AI Context Pack 证据范围，并建议用户先查看 Human Manual 或真实安装后验证。 - **runtime_request**：给出安装前检查清单和命令来源，但不要替用户执行命令或声称已执行。 - **source_conflict**：同时展示冲突来源，标记为待核实，不要强行选择一个版本。 ## Prompt Recipes ### 适配判断 - 目标：判断这个项目是否适合用户当前任务。 - 预期输出：适配结论、关键理由、证据引用、安装前可预览内容、必须安装后验证内容、下一步建议。 ```text 请基于 termyte 的 AI Context Pack，先问我 3 个必要问题，然后判断它是否适合我的任务。回答必须包含：适合谁、能做什么、不能做什么、是否值得安装、证据来自哪里。所有项目事实必须引用 evidence_refs、source_paths 或 claim_id。 ``` ### 安装前体验 - 目标：让用户在安装前感受核心工作流，同时避免把预览包装成真实能力或营销承诺。 - 预期输出：一段带边界标签的体验剧本、安装后验证清单和谨慎建议；不含真实运行承诺或强营销表述。 ```text 请把 termyte 当作安装前体验资产，而不是已安装工具或真实运行环境。 请严格输出四段： 1. 先问我 3 个必要问题。 2. 给出一段“体验剧本”：用 [安装前可预览]、[必须安装后验证]、[证据不足] 三种标签展示它可能如何引导工作流。 3. 给出安装后验证清单：列出哪些能力只有真实安装、真实宿主加载、真实项目运行后才能确认。 4. 给出谨慎建议：只能说“值得继续研究/试装”“先补充信息后再判断”或“不建议继续”，不得替项目背书。 硬性边界： - 不要声称已经安装、运行、执行测试、修改文件或产生真实结果。 - 不要写“自动适配”“确保通过”“完美适配”“强烈建议安装”等承诺性表达。 - 如果描述安装后的工作方式，必须使用“如果安装成功且宿主正确加载 Skill，它可能会……”这种条件句。 - 体验剧本只能写成“示例台词/假设流程”：使用“可能会询问/可能会建议/可能会展示”，不要写“已写入、已生成、已通过、正在运行、正在生成”。 - Prompt Preview 不负责给安装命令；如用户准备试装，只能提示先阅读 Quick Start 和 Risk Card，并在隔离环境验证。 - 所有项目事实必须来自 supported claim、evidence_refs 或 source_paths；inferred/unverified 只能作风险或待确认项。 ``` ### 角色 / Skill 选择 - 目标：从项目里的角色或 Skill 中挑选最匹配的资产。 - 预期输出：候选角色或 Skill 列表，每项包含适用场景、证据路径、风险边界和是否需要安装后验证。 ```text 请读取 role_skill_index，根据我的目标任务推荐 3-5 个最相关的角色或 Skill。每个推荐都要说明适用场景、可能输出、风险边界和 evidence_refs。 ``` ### 风险预检 - 目标：安装或引入前识别环境、权限、规则冲突和质量风险。 - 预期输出：环境、权限、依赖、许可、宿主冲突、质量风险和未知项的检查清单。 ```text 请基于 risk_card、boundaries 和 quick_start_candidates，给我一份安装前风险预检清单。不要替我执行命令，只说明我应该检查什么、为什么检查、失败会有什么影响。 ``` ### 宿主 AI 开工指令 - 目标：把项目上下文转成一次对话开始前的宿主 AI 指令。 - 预期输出：一段边界明确、证据引用明确、适合复制给宿主 AI 的开工前指令。 ```text 请基于 termyte 的 AI Context Pack，生成一段我可以粘贴给宿主 AI 的开工前指令。这段指令必须遵守 not_runtime=true，不能声称项目已经安装、运行或产生真实结果。 ``` ## 角色 / Skill 索引 - 共索引 10 个角色 / Skill / 项目文档条目。 - **Termyte Public Docs**（project_doc）：Termyte is a local execution and continuity layer for coding agents. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/README.md` - **Termyte**（project_doc）：Termyte helps coding agents remember what happened in a repository. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`README.md` - **Getting Started**（project_doc）：Install Termyte and connect Claude Code, Codex, or OpenCode: 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/getting-started.md` - **How It Works**（project_doc）：Termyte follows a simple runtime path: 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/how-it-works.md` - **Contributing**（project_doc）：Thanks for helping improve Termyte. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`CONTRIBUTING.md` - **Termyte Overview**（project_doc）：Termyte is a local execution, task-state, and memory layer for coding agents. It captures agent execution as an idempotent event ledger, maintains evidence-backed task state, derives reusable memories, and exposes continuity through CLI, MCP, context injection, handoffs, and a local diagnostics viewer. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`OVERVIEW.md` - **Comparisons**（project_doc）：This page compares Termyte with mem0, agentmemory, and claude-mem using the code and docs in the local repositories. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/comparisons.md` - **Limitations**（project_doc）：Termyte works, but it is not finished. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/limitations.md` - **Termyte Context Engine v0.1 Trial Protocol**（project_doc）：Termyte Context Engine v0.1 Trial Protocol 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/evals/context-v0.1-trial-protocol.md` - **Termyte Context Engine v0.1 Implementation Plan**（project_doc）：Termyte Context Engine v0.1 Implementation Plan 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/superpowers/plans/2026-07-14-termyte-context-engine-v0.1-implementation-plan.md` ## 证据索引 - 共索引 79 条证据。 - **Termyte Public Docs**（documentation）：Termyte is a local execution and continuity layer for coding agents. 证据：`docs/README.md` - **Termyte**（documentation）：Termyte helps coding agents remember what happened in a repository. 证据：`README.md` - **Package**（package_manifest）：{ "name": "termyte", "version": "1.0.5", "description": "Local-first context engine for coding agents", "repository": "https://github.com/termyte-labs/termyte", "type": "module", "license": "MIT", "engines": { "node": " =20" }, "main": "dist/index.js", "types": "dist/index.d.ts", "bin": { "termyte": "dist/cli/index.js", "termyte-hook": "dist/cli/hook.js", "termyte-worker": "dist/cli/worker.js" }, "files": "dist", "test/fixtures/regression-corpus/cases.json", "README.md", "LICENSE" , "scripts": { "build": "node -e \"require 'node:fs' .rmSync 'dist',{recursive:true,force:true} \" && tsc -p tsconfig.json && vite build", "typecheck": "tsc --noEmit -p tsconfig.json", "test": "vitest run --exclud… 证据：`package.json` - **Getting Started**（documentation）：Install Termyte and connect Claude Code, Codex, or OpenCode: 证据：`docs/getting-started.md` - **How It Works**（documentation）：Termyte follows a simple runtime path: 证据：`docs/how-it-works.md` - **Contributing**（documentation）：Thanks for helping improve Termyte. 证据：`CONTRIBUTING.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: 证据：`LICENSE` - **Termyte Overview**（documentation）：Termyte is a local execution, task-state, and memory layer for coding agents. It captures agent execution as an idempotent event ledger, maintains evidence-backed task state, derives reusable memories, and exposes continuity through CLI, MCP, context injection, handoffs, and a local diagnostics viewer. 证据：`OVERVIEW.md` - **Comparisons**（documentation）：This page compares Termyte with mem0, agentmemory, and claude-mem using the code and docs in the local repositories. 证据：`docs/comparisons.md` - **Limitations**（documentation）：Termyte works, but it is not finished. 证据：`docs/limitations.md` - **Termyte Context Engine v0.1 Trial Protocol**（documentation）：Termyte Context Engine v0.1 Trial Protocol 证据：`docs/evals/context-v0.1-trial-protocol.md` - **Termyte Context Engine v0.1 Implementation Plan**（documentation）：Termyte Context Engine v0.1 Implementation Plan 证据：`docs/superpowers/plans/2026-07-14-termyte-context-engine-v0.1-implementation-plan.md` - **Index**（source_file）：import { Store } from "./storage/store.js"; import { Observer } from "./observer/pipeline.js"; import { type OpenAIProviderConfig } from "./observer/openai-provider.js"; import { type LocalModelId } from "./retrieval/local-embeddings.js"; import { NoOpEmbeddingsProvider } from "./retrieval/embeddings.js"; import { FTSSearch } from "./retrieval/fts.js"; import { VectorSearch } from "./retrieval/vector.js"; import { HybridSearch } from "./retrieval/hybrid.js"; import { ContextBuilder } from "./context/builder.js"; import { HookRunner } from "./hooks/runner.js"; import { createEmbeddingsProvider, createLLMProvider } from "./runtime/providers.js"; ⋮---- export interface TermyteOptions { dbPath?… 证据：`src/index.ts` - **Termyte Fts**（source_file）：import { Store } from "../../storage/store.js"; import { FTSSearch } from "../../retrieval/fts.js"; import type { BenchmarkDocument, BenchmarkSearchResult, MemoryBenchmarkAdapter } from "../types.js"; ⋮---- export class TermyteFtsBenchmarkAdapter implements MemoryBenchmarkAdapter ⋮---- async reset : Promise ⋮---- async ingest documents: readonly BenchmarkDocument : Promise ⋮---- async search query: string, limit: number, options: ⋮---- async stats : Promise ⋮---- async close : Promise 证据：`src/benchmark/adapters/termyte-fts.ts` - **Termyte Hybrid**（source_file）：import { FTSSearch } from "../../retrieval/fts.js"; import { HybridSearch } from "../../retrieval/hybrid.js"; import { LocalEmbeddingsProvider, type LocalModelId } from "../../retrieval/local-embeddings.js"; import { VectorSearch } from "../../retrieval/vector.js"; import { Store } from "../../storage/store.js"; import type { BenchmarkDocument, BenchmarkSearchResult, MemoryBenchmarkAdapter } from "../types.js"; ⋮---- export class TermyteHybridBenchmarkAdapter implements MemoryBenchmarkAdapter ⋮---- constructor model: LocalModelId = "bge-small" ⋮---- async reset : Promise ⋮---- async ingest documents: readonly BenchmarkDocument : Promise ⋮---- async search query: string, limit: number, optio… 证据：`src/benchmark/adapters/termyte-hybrid.ts` - **Termyte Pipeline**（source_file）：import { FTSSearch } from "../../retrieval/fts.js"; import { FakeLLMProvider } from "../../observer/fake-provider.js"; import { MemoryPipeline } from "../../pipeline/memory-pipeline.js"; import { Store } from "../../storage/store.js"; import type { BenchmarkDocument, BenchmarkSearchResult, MemoryBenchmarkAdapter } from "../types.js"; ⋮---- export class TermytePipelineBenchmarkAdapter implements MemoryBenchmarkAdapter ⋮---- async reset : Promise ⋮---- async ingest documents: readonly BenchmarkDocument : Promise ⋮---- async search query: string, limit: number, options: ⋮---- async stats : Promise ⋮---- async close : Promise 证据：`src/benchmark/adapters/termyte-pipeline.ts` - **Metrics**（source_file）：import type { BenchmarkQuery, QueryEvaluation } from "./types.js"; ⋮---- export function evaluateQuery query: BenchmarkQuery, returnedDocumentIds: string , latencyMs: number, : QueryEvaluation ⋮---- export function aggregateEvaluations rows: readonly QueryEvaluation : Record ⋮---- const mean = pick: row: QueryEvaluation ⋮---- function ndcg returned: readonly string , relevant: ReadonlySet : number ⋮---- function percentile sorted: readonly number , fraction: number : number 证据：`src/benchmark/metrics.ts` - **Runner**（source_file）：import { createHash } from "node:crypto"; import { mkdir, readFile, writeFile } from "node:fs/promises"; import { join, resolve } from "node:path"; import { performance } from "node:perf hooks"; import { aggregateEvaluations, evaluateQuery } from "./metrics.js"; import type { BenchmarkDataset, BenchmarkTrack, MemoryBenchmarkAdapter, QueryEvaluation } from "./types.js"; ⋮---- export interface BenchmarkRunOptions { datasetPath?: string; dataset?: BenchmarkDataset; outputDirectory: string; adapter: MemoryBenchmarkAdapter; track: BenchmarkTrack; seed?: number; limit?: number; datasetLoader?: raw: string = BenchmarkDataset; } ⋮---- export async function runBenchmark options: BenchmarkRunOptions… 证据：`src/benchmark/runner.ts` - **Types**（source_file）：export type BenchmarkSuite = "memoryagent" "raw-session" "scale" "custom"; export type BenchmarkTrack = "retrieval" "pipeline"; ⋮---- export interface BenchmarkDocument { id: string; content: string; title?: string; files?: string ; metadata?: Record ; scope?: string; } ⋮---- export interface BenchmarkQuery { id: string; query: string; relevantDocumentIds: string ; harmfulDocumentIds?: string ; scope?: string; } ⋮---- export interface BenchmarkDataset { name: string; version: string; suite: BenchmarkSuite; documents: BenchmarkDocument ; queries: BenchmarkQuery ; } ⋮---- export interface BenchmarkSearchResult { documentId: string; score: number; } ⋮---- export interface MemoryBenchmarkAdapte… 证据：`src/benchmark/types.ts` - **Adapter**（source_file）：import type { EventType, Platform } from "../core/types.js"; ⋮---- export interface NormalizedEvent { session id: string; platform event id?: string null; timestamp: number; event type: EventType; tool name: string null; tool input: unknown null; tool output: unknown null; files read: string null; files modified: string null; user prompt: string null; final response: string null; cwd: string; } ⋮---- export interface HookResult { continue?: boolean; suppressOutput?: boolean; systemMessage?: string; hookSpecificOutput?: { hookEventName: string; additionalContext?: string; contextInjectionId?: string; permissionDecision?: "allow" "deny"; permissionDecisionReason?: string; updatedInput?: Recor… 证据：`src/capture/adapter.ts` - **Claude Code**（source_file）：import type { PlatformAdapter, NormalizedEvent, HookResult } from "./adapter.js"; import type { EventType } from "../core/types.js"; import { isObject, pickString } from "./util.js"; import { AdapterRejectedInput, isValidCwd } from "./errors.js"; ⋮---- const pickAgentField = v: unknown : string undefined ⋮---- export class ClaudeCodeAdapter implements PlatformAdapter ⋮---- normalize raw: unknown : NormalizedEvent null ⋮---- formatOutput result: HookResult : unknown 证据：`src/capture/claude-code.ts` - **Codex File Context**（source_file）：import { existsSync, statSync } from "node:fs"; import path from "node:path"; ⋮---- function flagsWithValues command: string : Set ⋮---- function dropFlagValue flag: string, command: string : boolean ⋮---- function isFlagLike value: string : boolean ⋮---- function isExistingFile candidate: string, cwd: string : boolean ⋮---- function tokenize command: string : string ⋮---- // Inside double quotes, only $, , ", \, and newline are escape-worthy. ⋮---- / Split a token list on ; , , && , so each pipeline is classified independently. / function splitSegments tokens: string : string ⋮---- // 2 &1 , & background, etc. — split, drop the operator. ⋮---- function normalizeCommand command: unknown : s… 证据：`src/capture/codex-file-context.ts` - **Codex**（source_file）：import type { PlatformAdapter, NormalizedEvent, HookResult } from "./adapter.js"; import type { EventType } from "../core/types.js"; import { isObject, pickString } from "./util.js"; import { extractCodexFilePaths } from "./codex-file-context.js"; import { AdapterRejectedInput, isValidCwd } from "./errors.js"; ⋮---- export class CodexAdapter implements PlatformAdapter ⋮---- normalize raw: unknown : NormalizedEvent null ⋮---- formatOutput result: HookResult : unknown 证据：`src/capture/codex.ts` - **Index**（source_file）：import type { Platform } from "../core/types.js"; import type { PlatformAdapter } from "./adapter.js"; import { ClaudeCodeAdapter } from "./claude-code.js"; import { CodexAdapter } from "./codex.js"; import { OpenCodeAdapter } from "./opencode.js"; import { RawAdapter } from "./raw.js"; ⋮---- export function adapterFor platform: Platform : PlatformAdapter 证据：`src/capture/index.ts` - **Opencode**（source_file）：import type { EventType } from "../core/types.js"; import type { HookResult, NormalizedEvent, PlatformAdapter } from "./adapter.js"; import { AdapterRejectedInput, isValidCwd } from "./errors.js"; import { isObject, pickString } from "./util.js"; ⋮---- export class OpenCodeAdapter implements PlatformAdapter ⋮---- normalize raw: unknown : NormalizedEvent null ⋮---- formatOutput result: HookResult : unknown 证据：`src/capture/opencode.ts` - **Doctor**（source_file）：import { existsSync, readFileSync } from "node:fs"; import { homedir } from "node:os"; import { join } from "node:path"; import { pathToFileURL } from "node:url"; import { loadConfig } from "./config.js"; import { isTermyteHookCommand } from "../integrations/installers/managed-hooks.js"; import { Store } from "../storage/store.js"; import { getTermyteHookPath } from "../integrations/install-paths.js"; ⋮---- type IntegrationStatus = { name: string; expectedPaths: string ; installed: boolean; evidence?: string; expectedHooks: string ; installedHooks: string ; missingHooks: string ; }; ⋮---- async function main : Promise ⋮---- export function inspectIntegrations : IntegrationStatus ⋮---- funct… 证据：`src/cli/doctor.ts` - **Handler Types**（source_file）：import type { HookResult, NormalizedEvent } from "../capture/adapter.js"; ⋮---- export interface HandlerInput { event: NormalizedEvent; raw: unknown; } ⋮---- export interface HandlerOutput { handled: boolean; result: HookResult; } ⋮---- export type EventHandlerName = "context" "session-init" "observation" "summarize" "file-edit" "file-context"; ⋮---- export type EventHandler = input: HandlerInput = Promise ; 证据：`src/cli/handler-types.ts` - **Index**（source_file）：import type { EventHandler, EventHandlerName } from "../handler-types.js"; import { makeContextHandler } from "./context.js"; import { makeSessionInitHandler } from "./session-init.js"; import { makeObservationHandler } from "./observation.js"; import { makeSummarizeHandler } from "./summarize.js"; import { makeFileEditHandler } from "./file-edit.js"; import { makeFileContextHandler } from "./file-context.js"; import type { Store } from "../../storage/store.js"; import type { HybridSearch } from "../../retrieval/hybrid.js"; import type { ContextBuilder } from "../../context/builder.js"; import type { Observer } from "../../observer/pipeline.js"; ⋮---- export interface HandlerDeps { store: S… 证据：`src/cli/handlers/index.ts` - **Session Init**（source_file）：import type { EventHandler } from "../handler-types.js"; import type { Store } from "../../storage/store.js"; import type { HybridSearch } from "../../retrieval/hybrid.js"; import type { ContextBuilder } from "../../context/builder.js"; import { makeContextHandler } from "./context.js"; ⋮---- export function makeSessionInitHandler deps: 证据：`src/cli/handlers/session-init.ts` - **Index**（source_file）：function parseOptions argv: string : Record ⋮---- async function main : Promise 证据：`src/cli/index.ts` - **Init**（source_file）：import { checkbox, confirm, select } from "@inquirer/prompts"; import { existsSync, mkdirSync, readFileSync, unlinkSync, writeFileSync } from "node:fs"; import { dirname, join } from "node:path"; import { homedir } from "node:os"; import { Store } from "../storage/store.js"; import { installFor } from "../integrations/installers/index.js"; import { discoverAgentCapabilities, capabilityLabel, type SupportedAgent } from "../runtime/capabilities.js"; import { removeTermyteHookEntries } from "../integrations/installers/managed-hooks.js"; import { openCodeConfigDir, uninstallOpenCode } from "../integrations/installers/opencode.js"; import { defaultUserConfig, saveUserConfig, userConfigPath, type… 证据：`src/cli/init.ts` - **Task**（source_file）：import { loadConfig } from "./config.js"; import { Store } from "../storage/store.js"; import { TaskStateService } from "../task-state/service.js"; import { CheckpointService } from "../task-state/checkpoints.js"; import { ResumeCompiler } from "../task-state/resume.js"; import type { Platform } from "../core/types.js"; ⋮---- type Options = Record ; ⋮---- export async function taskCommand action: string undefined, options: Options : Promise ⋮---- function required options: Options, name: string : string ⋮---- function print value: unknown : void 证据：`src/cli/task.ts` - **Uninstall**（source_file）：import { checkbox, confirm } from "@inquirer/prompts"; import { existsSync, rmSync } from "node:fs"; import { homedir } from "node:os"; import { join } from "node:path"; import { loadUserConfig, termyteHome } from "./config.js"; import { removeTermyteHookEntries } from "../integrations/installers/managed-hooks.js"; import { uninstallOpenCode } from "../integrations/installers/opencode.js"; ⋮---- type Removal = "claude-code" "codex" "opencode" "data"; ⋮---- export async function uninstallCommand : Promise ⋮---- export function uninstallTermyte selected: Removal , env: NodeJS.ProcessEnv = process.env : number 证据：`src/cli/uninstall.ts` - **Viewer**（source_file）：import { loadConfig } from "./config.js"; import { startViewerServer } from "../viewer/server.js"; import { spawn } from "node:child process"; ⋮---- export interface ViewerCommandOptions { host?: string; port?: number; open?: boolean; } ⋮---- export async function viewerCommand options: ViewerCommandOptions = ⋮---- function openBrowser url: string : void 证据：`src/cli/viewer.ts` - **Types**（source_file）：export type EventType = "session init" "user prompt" "tool use" "assistant message" "compaction" "session end"; ⋮---- export type Platform = "claude-code" "codex" "opencode" "raw"; ⋮---- export type MemoryType = "bugfix" "convention" "warning" "procedure" "fact"; ⋮---- export type ObservationType = MemoryType; ⋮---- export type MemoryState = "active" "stale" "superseded" "conflicted" "deleted"; ⋮---- export type MemoryEdgeType = "supports" "contradicts" "supersedes" "duplicates" "derived from" "related to"; ⋮---- export type MemoryFeedbackEvent = "shown" "used" "helpful" "harmful" "ignored" "downranked" "corrected"; ⋮---- export interface CodeApplicabilityEvidence { files: string ; commands… 证据：`src/core/types.ts` - **Fault Injection**（source_file）：import type { DB } from "../storage/connection.js"; ⋮---- export type FaultPoint = "after observation insert" "before observation embedding" "after observation embedding" "before memory insert" "after memory insert" "before memory embedding" "after memory embedding"; ⋮---- export class InjectedFaultError extends Error ⋮---- constructor readonly point: FaultPoint, message = Injected fault at $ ⋮---- export class FaultInjector ⋮---- failOnce point: FaultPoint : void ⋮---- alwaysFail point: FaultPoint : void ⋮---- clear point?: FaultPoint : void ⋮---- check point: FaultPoint : void ⋮---- export interface DurabilityInvariantReport { passed: boolean; violations: string ; } ⋮---- export function… 证据：`src/eval/fault-injection.ts` - **Harness**（source_file）：import { Store } from "../storage/store.js"; import { openDatabase } from "../storage/connection.js"; import { JobQueue } from "../pipeline/job-queue.js"; import { MemoryPipeline } from "../pipeline/memory-pipeline.js"; import { RetryableJobError } from "../pipeline/errors.js"; import { FTSSearch } from "../retrieval/fts.js"; import { VectorSearch } from "../retrieval/vector.js"; import { HybridSearch } from "../retrieval/hybrid.js"; import type { EmbeddingsProvider } from "../retrieval/embeddings.js"; import { memoryDecayScore } from "../lifecycle/decay.js"; import { applyFeedback } from "../lifecycle/feedback.js"; import { canonicalMemoryKey, shouldDeduplicate } from "../lifecycle/dedupe.… 证据：`src/eval/harness.ts` - **Metrics**（source_file）：export interface RankedResultLike { id: string; } ⋮---- export function recallAtK results: RankedResultLike , expectedIds: string , k: number : number ⋮---- export function precisionAtK results: RankedResultLike , expectedIds: string , k: number : number ⋮---- export function mrr results: RankedResultLike , expectedIds: string : number ⋮---- export function mean values: number : number ⋮---- function unique values: string : string 证据：`src/eval/metrics.ts` - **Runner**（source_file）：import type { Store } from "../storage/store.js"; import type { Observer } from "../observer/pipeline.js"; import type { Platform } from "../core/types.js"; import { adapterFor } from "../capture/index.js"; import { Ingestor } from "../capture/ingest.js"; import type { PlatformAdapter, NormalizedEvent, HookResult } from "../capture/adapter.js"; import { AdapterRejectedInput } from "../capture/errors.js"; import { detectRepoId, detectWorkspaceRoot } from "../retrieval/local-embeddings.js"; import { ExperienceRecorder } from "../experience/recorder.js"; ⋮---- export interface HookRunnerConfig { store: Store; observer: Observer; } ⋮---- export class HookRunner ⋮---- constructor config: HookRun… 证据：`src/hooks/runner.ts` - **Memory Vec Index**（source_file）：import { createRequire } from "node:module"; import type { DB } from "../storage/connection.js"; ⋮---- export interface MemoryVectorHit { memoryId: number; distance: number; score: number; } ⋮---- export class MemoryVecIndex ⋮---- constructor ⋮---- isAvailable : boolean tableName : string ⋮---- ensureSchema : boolean ⋮---- const sqliteVec = require "sqlite-vec" as ⋮---- upsert memoryId: number, embedding: Float32Array : void ⋮---- search query: Float32Array, limit: number : MemoryVectorHit ⋮---- private assertDimensions vector: Float32Array : void ⋮---- function vectorToBuffer vector: Float32Array : Buffer 证据：`src/indexing/memory-vec-index.ts` - **Sqlite Vec Index**（source_file）：import { createHash } from "node:crypto"; import { createRequire } from "node:module"; import type { DB } from "../storage/connection.js"; ⋮---- export interface VectorHit { docId: string; distance: number; score: number; } ⋮---- export interface SqliteVecIndexOptions { dimensions: number; model: string; forceUnavailable?: boolean; } ⋮---- export class SqliteVecIndex ⋮---- constructor ⋮---- isAvailable : boolean ⋮---- tableName : string ⋮---- ensureSchema : boolean ⋮---- upsert docId: string, vector: Float32Array : void ⋮---- search vector: Float32Array, limit: number : VectorHit ⋮---- private loadExtension : void ⋮---- const sqliteVec = require "sqlite-vec" as ⋮---- private assertAvailable… 证据：`src/indexing/sqlite-vec-index.ts` - **Claude Code**（source_file）：import { existsSync, readFileSync, writeFileSync, mkdirSync } from "node:fs"; import { dirname, join } from "node:path"; import { homedir } from "node:os"; import { getTermyteHookPath, shellEscapePath } from "../install-paths.js"; import { isTermyteHookCommand } from "./managed-hooks.js"; ⋮---- interface ClaudeHookEntry { type: "command"; command: string; timeout?: number; } ⋮---- interface ClaudeHooksGroup { matcher?: string; hooks: ClaudeHookEntry ; } ⋮---- interface ClaudeSettings { hooks?: Record ; k: string : unknown; } ⋮---- export interface ClaudeInstallOptions { target: "user" "project"; homeDir?: string; } ⋮---- export function installClaudeCodeHooks opts: ClaudeInstallOptions : nu… 证据：`src/integrations/installers/claude-code.ts` - **Codex**（source_file）：import { existsSync, readFileSync, writeFileSync, mkdirSync } from "node:fs"; import { dirname, join } from "node:path"; import { homedir } from "node:os"; import { getTermyteHookPath, shellEscapePath } from "../install-paths.js"; import { isTermyteHookCommand } from "./managed-hooks.js"; ⋮---- interface CodexHookEntry { type: "command"; command: string; timeout?: number; } interface CodexHookGroup { matcher?: string; hooks: CodexHookEntry ; } interface CodexHooksConfig { hooks?: Record ; k: string : unknown; } ⋮---- export interface CodexInstallOptions { target: "user" "project"; homeDir?: string; } ⋮---- export function installCodexHooks opts: CodexInstallOptions : number ⋮---- function s… 证据：`src/integrations/installers/codex.ts` - **Index**（source_file）：import { installClaudeCodeHooks } from "./claude-code.js"; import { installCodexHooks } from "./codex.js"; import { installOpenCode } from "./opencode.js"; ⋮---- export type SupportedPlatform = "claude-code" "codex" "opencode"; ⋮---- export function listSupportedPlatforms : string ⋮---- export interface InstallOptions { target?: "user" "project"; homeDir?: string; } ⋮---- export function installFor platform: string, opts: InstallOptions = 证据：`src/integrations/installers/index.ts` - **Opencode**（source_file）：import { existsSync, mkdirSync, readFileSync, rmSync, writeFileSync } from "node:fs"; import { join } from "node:path"; import { homedir } from "node:os"; ⋮---- type OpenCodeConfig = { plugin?: unknown; key: string : unknown }; ⋮---- export function openCodeConfigDir homeDir = homedir : string ⋮---- export function openCodePluginSource hookCommand = "termyte-hook" : string ⋮---- export function installOpenCode opts: ⋮---- export function uninstallOpenCode homeDir?: string : boolean 证据：`src/integrations/installers/opencode.ts` - **Server**（source_file）：import { loadConfig } from "../cli/config.js"; import { Store } from "../storage/store.js"; import { FTSSearch } from "../retrieval/fts.js"; import { VectorSearch } from "../retrieval/vector.js"; import { HybridSearch } from "../retrieval/hybrid.js"; import { ContextBuilder, renderHybridResults, renderMemory } from "../context/builder.js"; import { buildMemoryExplain, renderMemoryExplain } from "../explain/memory-explain.js"; import type { EmbeddingsProvider } from "../retrieval/embeddings.js"; import { NoOpEmbeddingsProvider } from "../retrieval/embeddings.js"; import { MCP TOOL DEFS } from "./tools.js"; import { DocumentStore, type DocumentType, type SparseHit } from "../storage/documents… 证据：`src/mcp/server.ts` - **Tools**（source_file）：import type { ToolDef } from "./types.js"; 证据：`src/mcp/tools.ts` - **Types**（source_file）：export interface JsonRpcRequest { jsonrpc: "2.0"; id?: number string; method: string; params?: unknown; } ⋮---- export interface JsonRpcResponse { jsonrpc: "2.0"; id: number string null; result?: unknown; error?: { code: number; message: string; data?: unknown }; } ⋮---- export interface ToolDef { name: string; description: string; inputSchema: Record ; } 证据：`src/mcp/types.ts` - **Agent Cli Provider**（source_file）：import type { LLMProvider, ChatMessage, ChatOptions, ChatResponse } from "./provider.js"; import { createAdapter } from "../synth/index.js"; import type { AgentAdapterId } from "../synth/types.js"; ⋮---- export class AgentCliLLMProvider implements LLMProvider ⋮---- constructor private readonly provider: Extract ⋮---- async chat messages: ChatMessage , options?: ChatOptions : Promise ⋮---- export class CaptureOnlyLLMProvider implements LLMProvider ⋮---- async chat : Promise 证据：`src/observer/agent-cli-provider.ts` - **Fake Provider**（source_file）：import { existsSync, writeFileSync } from "node:fs"; import type { ChatMessage, ChatOptions, ChatResponse, LLMProvider } from "./provider.js"; ⋮---- export class FakeLLMProvider implements LLMProvider ⋮---- async chat messages: ChatMessage , options?: ChatOptions : Promise ⋮---- private shouldFailOnce : boolean ⋮---- private buildObservation prompt: string : string ⋮---- private buildConsolidation prompt: string : string ⋮---- private buildSummary prompt: string : string ⋮---- function buildObservationXml input: { type: string; title: string; description: string; filesRead: string ; filesModified: string ; } : string ⋮---- function extractTag prompt: string, tag: string : string null ⋮----… 证据：`src/observer/fake-provider.ts` - **Openai Provider**（source_file）：import type { ChatMessage, ChatOptions, ChatResponse, LLMProvider } from "./provider.js"; ⋮---- export interface OpenAIProviderConfig { baseUrl: string; apiKey: string; model: string; defaultTemperature?: number; defaultMaxTokens?: number; } ⋮---- export class OpenAICompatibleProvider implements LLMProvider ⋮---- constructor private config: OpenAIProviderConfig ⋮---- async chat messages: ChatMessage , options?: ChatOptions : Promise 证据：`src/observer/openai-provider.ts` - **Parser**（source_file）：export interface ParsedObservation { type: string; title: string; description: string null; files read: string ; files modified: string ; } ⋮---- export interface ParsedSummary { summary text: string null; key changes: string ; key learnings: string ; skipped?: boolean; } ⋮---- export type ParseResult = { valid: true; observations: ParsedObservation ; summary: ParsedSummary null } { valid: false }; ⋮---- export function parseAgentXml raw: string : ParseResult ⋮---- function stripFences text: string : string ⋮---- function extractField content: string, fieldName: string : string null ⋮---- function extractArrayElements content: string, arrayName: string, elementName: string : string ⋮---- fu… 证据：`src/observer/parser.ts` - **Pipeline**（source_file）：import type { Store } from "../storage/store.js"; import type { Trace } from "../core/types.js"; import type { EmbeddingsProvider } from "../retrieval/embeddings.js"; import type { LLMProvider, ChatOptions } from "./provider.js"; import { JobQueue } from "../pipeline/job-queue.js"; import { MemoryPipeline } from "../pipeline/memory-pipeline.js"; ⋮---- export interface ObserverConfig { store: Store; llm: LLMProvider; embeddings?: EmbeddingsProvider; chatOptions?: ChatOptions; } ⋮---- export class Observer ⋮---- constructor config: ObserverConfig ⋮---- enqueue trace: Trace : void ⋮---- enqueueEpisode episodeId: string, nextRunAt?: number : void ⋮---- enqueueMany traces: Trace : void ⋮---- asy… 证据：`src/observer/pipeline.ts` - **Provider**（source_file）：export interface ChatMessage { role: "system" "user" "assistant"; content: string; } ⋮---- export interface ChatOptions { model?: string; temperature?: number; maxTokens?: number; signal?: AbortSignal; } ⋮---- export interface ChatResponse { content: string; model: string; usage?: { inputTokens: number; outputTokens: number; }; } ⋮---- export interface LLMProvider { chat messages: ChatMessage , options?: ChatOptions : Promise ; } ⋮---- chat messages: ChatMessage , options?: ChatOptions : Promise ; 证据：`src/observer/provider.ts` - **Job Queue**（source_file）：import type { DB } from "../storage/connection.js"; import { computeBackoffMs, isRetryableJobError, serializeJobError, } from "./errors.js"; ⋮---- export type JobState = "pending" "leased" "succeeded" "failed" "dead"; ⋮---- export type JobKind = "synthesize episode" "consolidate episode" "extract observation" "embed observation" "consolidate memory" "embed memory" "dedupe memories" "update summary" "decay memories" "verify memory" "attribute context"; ⋮---- export type JobSubjectType = "trace" "observation" "memory" "summary" "episode"; ⋮---- export interface Job { id: string; kind: JobKind; subjectType: JobSubjectType; subjectId: string; dedupeKey: string; state: JobState; attemptCount: nu… 证据：`src/pipeline/job-queue.ts` - **Memory Pipeline**（source_file）：import type { Store } from "../storage/store.js"; import { DocumentStore } from "../storage/documents.js"; import type { Trace, Observation, ObservationType, Memory, MemoryType } from "../core/types.js"; import type { EmbeddingsProvider } from "../retrieval/embeddings.js"; import type { LLMProvider, ChatOptions } from "../observer/provider.js"; import { parseAgentXml } from "../observer/parser.js"; import { buildConsolidationPrompt, buildConsolidationSystemPrompt, buildObservationBatchPrompt, buildObservationPrompt, buildSummaryPrompt, buildSummarySystemPrompt, buildSystemPrompt, type SessionForPrompt, } from "../observer/prompts.js"; import { chatWithRetry } from "../observer/self-correct.… 证据：`src/pipeline/memory-pipeline.ts` - **Worker Supervisor**（source_file）：import { spawn } from "node:child process"; import { closeSync, existsSync, openSync, readFileSync, unlinkSync, writeFileSync, } from "node:fs"; import { dirname, join, resolve } from "node:path"; import { fileURLToPath } from "node:url"; ⋮---- export interface WorkerSupervisor { maybeLaunch : boolean; } ⋮---- maybeLaunch : boolean; ⋮---- export interface DetachedSupervisorConfig { dbPath: string; workerPath: string; nodeExecutable?: string; enabled: boolean; env?: NodeJS.ProcessEnv; } ⋮---- export function resolveWorkerPath env: NodeJS.ProcessEnv = process.env : string null ⋮---- export class DetachedWorkerSupervisor implements WorkerSupervisor ⋮---- constructor cfg: DetachedSupervisorConf… 证据：`src/pipeline/worker-supervisor.ts` - **Fts**（source_file）：import type { Store } from "../storage/store.js"; import type { Memory, MemoryType } from "../core/types.js"; import { memoryEligibilitySql } from "./eligibility.js"; import { preprocessQuery } from "./query-preprocessor.js"; ⋮---- export interface FTSSearchOptions { query: string; repo id?: string; limit?: number; type?: string; eligibleStates?: readonly string ; } ⋮---- export class FTSSearch ⋮---- constructor private store: Store ⋮---- search options: FTSSearchOptions : Memory ⋮---- function buildFTSQuery query: string : string ⋮---- function mapMemoryRow row: any : Memory ⋮---- function parseJSON s: string null undefined, fallback: T : T 证据：`src/retrieval/fts.ts` - **Hybrid**（source_file）：import type { Memory } from "../core/types.js"; import type { EmbeddingsProvider } from "./embeddings.js"; import { FTSSearch, type FTSSearchOptions } from "./fts.js"; import { VectorSearch, type VectorSearchOptions, type VectorSearchResult } from "./vector.js"; import { scoreMemoryCandidate, type RetrievalScoreBreakdown } from "./ranking.js"; import { CrossEncoderReranker } from "./reranker.js"; ⋮---- export interface HybridSearchOptions { query: string; repo id?: string; limit?: number; type?: string; recencyWindowMs?: number; currentFiles?: string ; eligibleStates?: readonly string ; } ⋮---- export interface HybridSearchResult { memory: Memory; fts rank?: number; vector rank?: number; co… 证据：`src/retrieval/hybrid.ts` - **Ranking**（source_file）：import type { Memory } from "../core/types.js"; ⋮---- export interface RetrievalScoreBreakdown { fts rrf: number; vector rrf: number; base score: number; confidence adjustment: number; importance adjustment: number; decay adjustment: number; usage adjustment: number; feedback adjustment: number; applicability adjustment: number; multiplier: number; final score: number; } ⋮---- export function scoreMemoryCandidate input: { memory: Memory; ftsRank?: number; vectorRank?: number; feedbackScore?: number; query?: string; currentFiles?: string ; rrfK?: number; } : RetrievalScoreBreakdown ⋮---- function clamp01 value: number : number function clamp value: number, min: number, max: number : number ⋮… 证据：`src/retrieval/ranking.ts` - 其余 19 条证据见 `AI_CONTEXT_PACK.json` 或 `EVIDENCE_INDEX.json`。 ## 宿主 AI 必须遵守的规则 - **把本资产当作开工前上下文，而不是运行环境。**：AI Context Pack 只包含证据化项目理解，不包含目标项目的可执行状态。 证据：`docs/README.md`, `README.md`, `package.json` - **回答用户时区分可预览内容与必须安装后才能验证的内容。**：安装前体验的消费者价值来自降低误装和误判，而不是伪装成真实运行。 证据：`docs/README.md`, `README.md`, `package.json` ## 用户开工前应该回答的问题 - 你准备在哪个宿主 AI 或本地环境中使用它？ - 你只是想先体验工作流，还是准备真实安装？ - 你最在意的是安装成本、输出质量、还是和现有规则的冲突？ ## 验收标准 - 所有能力声明都能回指到 evidence_refs 中的文件路径。 - AI_CONTEXT_PACK.md 没有把预览包装成真实运行。 - 用户能在 3 分钟内看懂适合谁、能做什么、如何开始和风险边界。 --- ## Doramagic Context Augmentation 下面内容用于强化 Repomix/AI Context Pack 主体。Human Manual 只提供阅读骨架；踩坑日志会被转成宿主 AI 必须遵守的工作约束。 ## Human Manual 骨架 使用规则：这里只是项目阅读路线和显著性信号，不是事实权威。具体事实仍必须回到 repo evidence / Claim Graph。 宿主 AI 硬性规则： - 不得把页标题、章节顺序、摘要或 importance 当作项目事实证据。 - 解释 Human Manual 骨架时，必须明确说它只是阅读路线/显著性信号。 - 能力、安装、兼容性、运行状态和风险判断必须引用 repo evidence、source path 或 Claim Graph。 - **项目概览与快速上手**：importance `high` - source_paths: README.md, OVERVIEW.md, docs/getting-started.md, docs/how-it-works.md - **系统整体架构**：importance `high` - source_paths: src/index.ts, src/core/types.ts, src/pipeline/memory-pipeline.ts, src/pipeline/job-queue.ts, src/pipeline/worker-supervisor.ts - **代理事件采集与适配器**：importance `high` - source_paths: src/capture/index.ts, src/capture/claude-code.ts, src/capture/codex.ts, src/capture/codex-file-context.ts, src/capture/opencode.ts - **事件观察者与 LLM 处理管线**：importance `high` - source_paths: src/observer/pipeline.ts, src/observer/parser.ts, src/observer/openai-provider.ts, src/observer/agent-cli-provider.ts, src/observer/fake-provider.ts - **检索、嵌入与索引**：importance `high` - source_paths: src/retrieval/hybrid.ts, src/retrieval/rrf.ts, src/retrieval/ranking.ts, src/retrieval/reranker.ts, src/retrieval/fts.ts - **记忆合成与生命周期管理**：importance `high` - source_paths: src/synth/index.ts, src/synth/resolve.ts, src/synth/claude-code.ts, src/synth/codex.ts, src/synth/opencode.ts - **任务状态、检查点与恢复**：importance `medium` - source_paths: src/task-state/service.ts, src/task-state/checkpoints.ts, src/task-state/resume.ts - **存储层与数据管理**：importance `high` - source_paths: src/storage/connection.ts, src/storage/documents.ts, src/storage/migrations.ts ## Repo Inspection Evidence / 源码检查证据 - repo_clone_verified: true - repo_inspection_verified: true - repo_commit: `70665e3c933b451fddce3cd6b820d2306e876f14` - inspected_files: `README.md`, `package.json`, `docs/README.md`, `docs/comparisons.md`, `docs/evals/context-v0.1-trial-protocol.md`, `docs/getting-started.md`, `docs/how-it-works.md`, `docs/limitations.md`, `docs/superpowers/plans/2026-07-14-termyte-context-engine-v0.1-implementation-plan.md`, `docs/superpowers/specs/2026-07-11-termyte-context-engine-v0.1-design.md`, `docs/superpowers/specs/2026-07-18-opencode-synthesis-adapter-design.md`, `src/benchmark/adapters/grep.ts`, `src/benchmark/adapters/termyte-fts.ts`, `src/benchmark/adapters/termyte-hybrid.ts`, `src/benchmark/adapters/termyte-pipeline.ts`, `src/benchmark/comparison.ts`, `src/benchmark/datasets/memoryagentbench.ts`, `src/benchmark/datasets/raw-session.ts`, `src/benchmark/datasets/scale.ts`, `src/benchmark/metrics.ts` 宿主 AI 硬性规则： - 没有 repo_clone_verified=true 时，不得声称已经读过源码。 - 没有 repo_inspection_verified=true 时，不得把 README/docs/package 文件判断写成事实。 - 没有 quick_start_verified=true 时，不得声称 Quick Start 已跑通。 ## Doramagic Pitfall Constraints / 踩坑约束 这些规则来自 Doramagic 发现、验证或编译过程中的项目专属坑点。宿主 AI 必须把它们当作工作约束，而不是普通说明文字。 ### Constraint 1: 可能修改宿主 AI 配置 - Trigger: 项目面向 Claude/Cursor/Codex/Gemini/OpenCode 等宿主，或安装命令涉及用户配置目录。 - Host AI rule: 列出会写入的配置文件、目录和卸载/回滚步骤。 - Why it matters: 安装可能改变本机 AI 工具行为，用户需要知道写入位置和回滚方法。 - Evidence: capability.host_targets | https://github.com/termyte-labs/termyte | host_targets=mcp_host, claude_code, claude, chatgpt - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 2: 能力判断依赖假设 - Trigger: README/documentation is current enough for a first validation pass. - Host AI rule: 将假设转成下游验证清单。 - Why it matters: 假设不成立时，用户拿不到承诺的能力。 - Evidence: capability.assumptions | https://github.com/termyte-labs/termyte | README/documentation is current enough for a first validation pass. - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 3: 维护活跃度未知 - Trigger: 未记录 last_activity_observed。 - Host AI rule: 补 GitHub 最近 commit、release、issue/PR 响应信号。 - Why it matters: 新项目、停更项目和活跃项目会被混在一起，推荐信任度下降。 - Evidence: evidence.maintainer_signals | https://github.com/termyte-labs/termyte | last_activity_observed missing - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 - Trigger: no_demo - Evidence: downstream_validation.risk_items | https://github.com/termyte-labs/termyte | no_demo; severity=medium - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 5: 存在评分风险 - Trigger: no_demo - Why it matters: 风险会影响是否适合普通用户安装。 - Evidence: risks.scoring_risks | https://github.com/termyte-labs/termyte | no_demo; severity=medium - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 6: issue/PR 响应质量未知 - Trigger: issue_or_pr_quality=unknown。 - Host AI rule: 抽样最近 issue/PR，判断是否长期无人处理。 - Why it matters: 用户无法判断遇到问题后是否有人维护。 - Evidence: evidence.maintainer_signals | https://github.com/termyte-labs/termyte | issue_or_pr_quality=unknown - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。 ### Constraint 7: 发布节奏不明确 - Trigger: release_recency=unknown。 - Host AI rule: 确认最近 release/tag 和 README 安装命令是否一致。 - Why it matters: 安装命令和文档可能落后于代码，用户踩坑概率升高。 - Evidence: evidence.maintainer_signals | https://github.com/termyte-labs/termyte | release_recency=unknown - Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。