[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7AVMK0RWQC0Y74PDVVMW83J":3,"skill-same:01M7AVMK0RWQC0Y74PDVVMW83J":77},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":29,"updatedAt":30,"install":31,"resources":44},"01M7AVMK0RWQC0Y74PDVVMW83J","codebase-context-architect",null,"Architect, synchronize, budget, and verify multi-layer codebase context files (AGENTS.md, CLAUDE.md, .cursor\u002Frules\u002F). Implement single-source generators, automated CI linters, and post-tool hooks to eliminate context rot. Do not use for generic prompt writing or marketing documentation.","SpectreDeath\u002FBrain-Harness","SpectreDeath","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F207463416?v=4","01M7AVGR1D4X8ADT47APKGVH2Z","JL","01M7AVGR2HSGD23BA5DBTJJTVN",0,"MIT","https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: codebase-context-architect\ndescription: Architect, synchronize, budget, and verify multi-layer codebase context files (AGENTS.md, CLAUDE.md, .cursor\u002Frules\u002F). Implement single-source generators, automated CI linters, and post-tool hooks to eliminate context rot. Do not use for generic prompt writing or marketing documentation.\n---\n\n# Codebase Context Architect\n\n`codebase-context-architect` is the repository engineering and context governance engine for managing AI agent context files across complex codebases. It operationalizes foundational principles from Kayode Adeniyi (*How to Manage Context Files in Your Codebase and Get Better Output From AI Coding Agents*, freeCodeCamp \u002F MCF) and Anthropic's context engineering research.\n\nIt eliminates **Context Rot**, **Kitchen Sink Bloat**, and **Vendor Fork Divergence** by structuring repository instructions into a **Three-Layer Context Architecture**, synchronizing multi-tool formats from a canonical Single Source of Truth (`AGENTS.md`), and enforcing automated CI verification linters that fail builds on broken paths, dead scripts, and budget blowouts.\n\nSee [CARD.md](CARD.md) for the companion summary card, 5-stage progression table, and invariants checklist.\nConsult `\u002Fagent-skills-architect` for enterprise skill design, `\u002Fagent-skill-sdlc` for lifecycle management, `\u002Fagent-instruction-architect` for instruction smell auditing, [three-layers-architecture.md](references\u002Fthree-layers-architecture.md) for architectural partitioning, and [context-linter-guide.md](references\u002Fcontext-linter-guide.md) for automated CI verification.\n\n---\n\n## The 5-Stage Context Architecture Progression\n\n```\n[1. Layered Partitioning & Removal Test] ──► [2. SSOT Generation & Vendor Sync] ──► [3. Altitude & Deliberate Absences]\n                                                                                               │\n                                                                                               ▼\n[5. Empirical Benchmarks & DoD Feedback] ◄── [4. Automated Context Verification & CI] ◄────────┘\n```\n\n---\n\n## 1. Layered Context Partitioning & Removal Test\n\nTreat context files as a **finite budget allocation problem** before a documentation problem:\n\n1. **Partition Context into Three Discrete Layers**:\n   - **Layer 1: Always Loaded Root (`AGENTS.md`)**:\n     - Read at the start of every session regardless of task.\n     - Budget ceiling: strictly bounded to $\\le 800$ tokens (~3,200 characters).\n     - Contains only repository-wide invariants, essential development loop commands, and non-obvious conventions.\n   - **Layer 2: Scoped Subdirectory Files (`src\u002FAGENTS.md`, `.cursor\u002Frules\u002F*.mdc`)**:\n     - Loads only when the agent operates within a specific subdirectory or touches matching file globs.\n     - Budget ceiling: strictly bounded to $\\le 400$ tokens (~1,600 characters).\n     - Houses domain-specific rules (e.g., API handler contracts, frontend state stores, test isolation).\n   - **Layer 3: On-Demand Pointers (`docs\u002F`, `skills\u002F`)**:\n     - Costs only a few tokens in the root file as path references (`docs\u002Farchitecture.md`, `docs\u002Ftesting.md`, `docs\u002Fdecisions\u002F`).\n     - Agents pull the underlying detailed documentation into context only when a specific task requires it.\n2. **Apply the Strict Line Removal Test**:\n   - For every line considered for inclusion, ask:\n     > *\"Would removing this line cause the agent to make a mistake?\"*\n   - If the answer is **no**, purge the line immediately.\n   - Eliminate basic language explanations (e.g. `const` vs `var`), directory walk descriptions visible from the tree, and generic platitudes.\n\n> **Completion criterion**: Context partitioned across 3 layers with always-loaded files under token budgets and non-essential prose pruned.\n\n---\n\n## 2. SSOT Generation & Vendor Synchronization\n\nEliminate vendor divergence by declaring canonical `AGENTS.md` as the sole human-edited source of truth:\n\n1. **Establish the Canonical Source of Truth**:\n   - Author all root instructions in `AGENTS.md` adhering to the Linux Foundation \u002F Agentic AI Foundation standard.\n2. **Automate Downstream Vendor File Generation**:\n   - Use `harness context sync` (or `python scripts\u002Fsync_context.py` \u002F `CodebaseContextEngine.sync()`):\n     - **Claude Code (`CLAUDE.md`)**:\n       - Leverage Claude Code's `@` file import syntax (`@AGENTS.md`).\n       - Append only tool-specific operational rules (`CLAUDE_EXTRAS`, e.g., plan mode triggers, subagent delegation preferences).\n       - Keeps `CLAUDE.md` compact ($\\le 150$ tokens).\n     - **GitHub Copilot (`.github\u002Fcopilot-instructions.md`)**:\n       - Inlines canonical `AGENTS.md` content with a prominent generated banner.\n     - **Cursor (`.cursor\u002Frules\u002F*.mdc`)**:\n       - Generate glob-scoped `.mdc` files with YAML frontmatter specifying directory match globs (`globs: \"src\u002F**\u002F*.ts\"`).\n3. **Embed Anti-Tamper Warning Banners**:\n   - Inject a visible header comment into all generated files:\n     ```markdown\n     \u003C!-- Generated from AGENTS.md by `harness context sync`. Edit AGENTS.md instead. -->\n     ```\n   - Wire CI checks to fail if generated files are edited directly without running the synchronizer (`harness context sync --dry-run`).\n\n\n> **Completion criterion**: Canonical AGENTS.md authored, downstream generator script configured, and vendor files synchronized with zero manual duplication.\n\n---\n\n## 3. Altitude Calibration & Deliberate Absences\n\nCalibrate instruction altitude and protect architectural choices from speculative agent additions:\n\n1. **Calibrate Instruction Altitude**:\n   - Avoid **Rigid Brittleness** (line-level micro-rules that shatter on the first valid exception).\n   - Avoid **Vague Encouragement** (\"write clean, maintainable code\" which changes zero agent decisions).\n   - Follow the **Three-Part Altitude Formula**:\n     1. **Pattern Shape**: Describe the structural role of the module (e.g., *\"Handlers parse and validate input, then delegate to services\"*).\n     2. **Explicit Boundary**: State the hard negative boundary (e.g., *\"Handlers never touch the database store directly\"*).\n     3. **Worked Example Link**: Point to an existing file demonstrating the pattern (e.g., *\"See `src\u002Fapi\u002Ftasks.js` for the pattern to copy\"*).\n2. **Attach Positive Rationales to Rules**:\n   - State the reason behind non-obvious constraints (*\"Handers return error codes rather than HTTP statuses because `router.js` owns status mapping\"*).\n   - Rules with attached reasons survive novel situations unpredicted by the rule author.\n3. **Document Deliberate Absences in ADRs**:\n   - Capture deliberate omissions in Architecture Decision Records (`docs\u002Fdecisions\u002F`).\n   - Explicitly instruct the agent that missing infrastructure is an intentional design choice rather than a gap to fill (e.g., *\"This service uses an in-memory store by design; do not introduce Postgres, SQLite, or an ORM\"*).\n\n> **Completion criterion**: Instructions calibrated with pattern shapes, negative boundaries, worked examples, and ADR intentional absences documented.\n\n---\n\n## 4. Automated Context Verification & CI Gating\n\nPrevent silent context rot by treating context files like executable code verified by automated linters:\n\n1. **Execute the 4-Check Context Linter (`harness context lint` \u002F `CodebaseContextEngine.lint()`)**:\n   - **Check 1: Token Budget Ceilings**:\n     - Estimate token counts across always-loaded files using calibrated character heuristics.\n     - Flag any file approaching or exceeding its budget ceiling.\n   - **Check 2: Prose Inline Path Integrity**:\n     - Strip fenced markdown codeblocks (` ```...``` `).\n     - Extract all backtick inline code spans resembling paths (` `src\u002Fservices\u002Ftasks.py` `).\n     - Assert that every referenced path exists on disk (`Path.exists()`).\n   - **Check 3: Script Command Validity**:\n     - Extract all mentioned script invocations (e.g., `npm test`, `pytest`, `ruff check`).\n     - Verify against declared scripts in `package.json`, `pyproject.toml`, or `Makefile`.\n   - **Check 4: Generator Sync Drift**:\n     - Re-run `harness context sync --dry-run` or `sync_context.py --dry-run`.\n     - Fail if on-disk vendor files differ by even a single character from canonical generation.\n2. **Wire into CI Pipelines & Git Hooks**:\n   - Add context linting to GitHub Actions \u002F CI workflow:\n     ```yaml\n     - name: Verify Codebase Context Files\n       run: harness context lint --root .\n     ```\n   - Wire pre-commit hooks to block commits containing broken paths or drifted vendor files.\n\n\n> **Completion criterion**: Context linter passes with exit code 0, 0 broken paths, 0 missing scripts, and zero sync drift.\n\n---\n\n## 5. Empirical Benchmarking & Definition of Done Feedback\n\nGround agent execution with deterministic feedback loops and objective validation:\n\n1. **Formulate Verifiable Definitions of Done**:\n   - Never allow an agent to declare a task finished based on \"looks completed\".\n   - Mandate specific test\u002Flint execution commands in the root file.\n   - **Require Evidence**: Demand that the agent paste the actual terminal output rather than claiming success in prose.\n2. **Install PostToolUse Agent Hooks**:\n   - Where supported by the client harness (e.g. Claude Code `.claude\u002Fsettings.json`), install `PostToolUse` hooks on `Edit|Write` actions that silently run context linting:\n     ```json\n     {\n       \"hooks\": {\n         \"PostToolUse\": [\n           {\n             \"matcher\": \"Edit|Write\",\n             \"hooks\": [\n               {\n                 \"type\": \"command\",\n                 \"command\": \"python scripts\u002Fcontext_linter.py --silent\"\n               }\n             ]\n           }\n         ]\n       }\n     }\n     ```\n3. **Execute Empirical A\u002FB Benchmarks**:\n   - Validate context file ROI by testing a standardized feature implementation prompt under two conditions:\n     - **Branch A**: With full multi-layer context files active.\n     - **Branch B**: With context files deleted (baseline).\n   - Score the delta along 4 objective metrics:\n     1. Test pass rate on first attempt.\n     2. Number of human interventions\u002Fcorrections required.\n     3. Adherence to architectural boundaries and layering.\n     4. Zero unapproved third-party dependencies introduced.\n\n> **Completion criterion**: Definition of done enforced with pasted evidence, post-tool hooks wired, and empirical A\u002FB validation completed.\n\n---\n\n## The Three Foundational Pillars\n\n### 1. The Visual Brief Pillar\nEvery major context restructuring or repository onboarding session must synthesize the context topology into an interactive HTML visual brief in `%TEMP%` loading Tailwind CSS and Mermaid.js via CDN. The brief visualizes the 3 layers, token budgets, and linter status.\n\n### 2. The Mandatory Checkpoint Gate Pillar\nWhen re-architecting repository context files, partitioning rules, or configuring CI gates, author an `implementation_plan.md` artifact with `RequestFeedback: true`. The agent must **STOP and wait** for explicit developer review and approval before mutating workspace files.\n\n### 3. Explicit Anti-Patterns\nRigid behavioral boundaries must be maintained to prevent context rot, instruction pollution, and ungrounded agent drift.\n\n---\n\n## Anti-Patterns\n\n- **Kitchen Sink Bloat** — Inlining generic tutorials, language default lessons, or team history into root files, exhausting model attention budgets.\n- **Silent Context Rot** — Allowing stale backtick paths and deleted scripts to linger because no CI check or pre-commit hook verifies them.\n- **Vendor Fork Divergence** — Maintaining separate hand-edited copies of CLAUDE.md, AGENTS.md, and Copilot instructions that contradict each other.\n- **Vague Encouragement Trap** — Stating ungrounded platitudes (\"write maintainable code\") instead of concrete boundaries, patterns, and worked examples.\n- **Inlined Monolith Sprawl** — Inlining multi-page architecture guides and schemas into root instructions instead of pointing to on-demand docs.\n- **Unverifiable Looks-Done Exit** — Allowing agents to finish tasks without running concrete test\u002Flint commands and pasting verifiable CLI output evidence.\n","# Codebase Context Architect\n\n`codebase-context-architect` is the repository engineering and context governance engine for managing AI agent context files across complex codebases. It operationalizes foundational principles from Kayode Adeniyi (*How to Manage Context Files in Your Codebase and Get Better Output From AI Coding Agents*, freeCodeCamp \u002F MCF) and Anthropic's context engineering research.\n\nIt eliminates **Context Rot**, **Kitchen Sink Bloat**, and **Vendor Fork Divergence** by structuring repository instructions into a **Three-Layer Context Architecture**, synchronizing multi-tool formats from a canonical Single Source of Truth (`AGENTS.md`), and enforcing automated CI verification linters that fail builds on broken paths, dead scripts, and budget blowouts.\n\nSee [CARD.md](CARD.md) for the companion summary card, 5-stage progression table, and invariants checklist.\nConsult `\u002Fagent-skills-architect` for enterprise skill design, `\u002Fagent-skill-sdlc` for lifecycle management, `\u002Fagent-instruction-architect` for instruction smell auditing, [three-layers-architecture.md](references\u002Fthree-layers-architecture.md) for architectural partitioning, and [context-linter-guide.md](references\u002Fcontext-linter-guide.md) for automated CI verification.\n\n---\n\n## The 5-Stage Context Architecture Progression\n\n```\n[1. Layered Partitioning & Removal Test] ──► [2. SSOT Generation & Vendor Sync] ──► [3. Altitude & Deliberate Absences]\n                                                                                               │\n                                                                                               ▼\n[5. Empirical Benchmarks & DoD Feedback] ◄── [4. Automated Context Verification & CI] ◄────────┘\n```\n\n---\n\n## 1. Layered Context Partitioning & Removal Test\n\nTreat context files as a **finite budget allocation problem** before a documentation problem:\n\n1. **Partition Context into Three Discrete Layers**:\n   - **Layer 1: Always Loaded Root (`AGENTS.md`)**:\n     - Read at the start of every session regardless of task.\n     - Budget ceiling: strictly bounded to $\\le 800$ tokens (~3,200 characters).\n     - Contains only repository-wide invariants, essential development loop commands, and non-obvious conventions.\n   - **Layer 2: Scoped Subdirectory Files (`src\u002FAGENTS.md`, `.cursor\u002Frules\u002F*.mdc`)**:\n     - Loads only when the agent operates within a specific subdirectory or touches matching file globs.\n     - Budget ceiling: strictly bounded to $\\le 400$ tokens (~1,600 characters).\n     - Houses domain-specific rules (e.g., API handler contracts, frontend state stores, test isolation).\n   - **Layer 3: On-Demand Pointers (`docs\u002F`, `skills\u002F`)**:\n     - Costs only a few tokens in the root file as path references (`docs\u002Farchitecture.md`, `docs\u002Ftesting.md`, `docs\u002Fdecisions\u002F`).\n     - Agents pull the underlying detailed documentation into context only when a specific task requires it.\n2. **Apply the Strict Line Removal Test**:\n   - For every line considered for inclusion, ask:\n     > *\"Would removing this line cause the agent to make a mistake?\"*\n   - If the answer is **no**, purge the line immediately.\n   - Eliminate basic language explanations (e.g. `const` vs `var`), directory walk descriptions visible from the tree, and generic platitudes.\n\n> **Completion criterion**: Context partitioned across 3 layers with always-loaded files under token budgets and non-essential prose pruned.\n\n---\n\n## 2. SSOT Generation & Vendor Synchronization\n\nEliminate vendor divergence by declaring canonical `AGENTS.md` as the sole human-edited source of truth:\n\n1. **Establish the Canonical Source of Truth**:\n   - Author all root instructions in `AGENTS.md` adhering to the Linux Foundation \u002F Agentic AI Foundation standard.\n2. **Automate Downstream Vendor File Generation**:\n   - Use `harness context sync` (or `python scripts\u002Fsync_context.py` \u002F `CodebaseContextEngine.sync()`):\n     - **Claude Code (`CLAUDE.md`)**:\n       - Leverage Claude Code's `@` file import syntax (`@AGENTS.md`).\n       - Append only tool-specific operational rules (`CLAUDE_EXTRAS`, e.g., plan mode triggers, subagent delegation preferences).\n       - Keeps `CLAUDE.md` compact ($\\le 150$ tokens).\n     - **GitHub Copilot (`.github\u002Fcopilot-instructions.md`)**:\n       - Inlines canonical `AGENTS.md` content with a prominent generated banner.\n     - **Cursor (`.cursor\u002Frules\u002F*.mdc`)**:\n       - Generate glob-scoped `.mdc` files with YAML frontmatter specifying directory match globs (`globs: \"src\u002F**\u002F*.ts\"`).\n3. **Embed Anti-Tamper Warning Banners**:\n   - Inject a visible header comment into all generated files:\n     ```markdown\n     \u003C!-- Generated from AGENTS.md by `harness context sync`. Edit AGENTS.md instead. -->\n     ```\n   - Wire CI checks to fail if generated files are edited directly without running the synchronizer (`harness context sync --dry-run`).\n\n\n> **Completion criterion**: Canonical AGENTS.md authored, downstream generator script configured, and vendor files synchronized with zero manual duplication.\n\n---\n\n## 3. Altitude Calibration & Deliberate Absences\n\nCalibrate instruction altitude and protect architectural choices from speculative agent additions:\n\n1. **Calibrate Instruction Altitude**:\n   - Avoid **Rigid Brittleness** (line-level micro-rules that shatter on the first valid exception).\n   - Avoid **Vague Encouragement** (\"write clean, maintainable code\" which changes zero agent decisions).\n   - Follow the **Three-Part Altitude Formula**:\n     1. **Pattern Shape**: Describe the structural role of the module (e.g., *\"Handlers parse and validate input, then delegate to services\"*).\n     2. **Explicit Boundary**: State the hard negative boundary (e.g., *\"Handlers never touch the database store directly\"*).\n     3. **Worked Example Link**: Point to an existing file demonstrating the pattern (e.g., *\"See `src\u002Fapi\u002Ftasks.js` for the pattern to copy\"*).\n2. **Attach Positive Rationales to Rules**:\n   - State the reason behind non-obvious constraints (*\"Handers return error codes rather than HTTP statuses because `router.js` owns status mapping\"*).\n   - Rules with attached reasons survive novel situations unpredicted by the rule author.\n3. **Document Deliberate Absences in ADRs**:\n   - Capture deliberate omissions in Architecture Decision Records (`docs\u002Fdecisions\u002F`).\n   - Explicitly instruct the agent that missing infrastructure is an intentional design choice rather than a gap to fill (e.g., *\"This service uses an in-memory store by design; do not introduce Postgres, SQLite, or an ORM\"*).\n\n> **Completion criterion**: Instructions calibrated with pattern shapes, negative boundaries, worked examples, and ADR intentional absences documented.\n\n---\n\n## 4. Automated Context Verification & CI Gating\n\nPrevent silent context rot by treating context files like executable code verified by automated linters:\n\n1. **Execute the 4-Check Context Linter (`harness context lint` \u002F `CodebaseContextEngine.lint()`)**:\n   - **Check 1: Token Budget Ceilings**:\n     - Estimate token counts across always-loaded files using calibrated character heuristics.\n     - Flag any file approaching or exceeding its budget ceiling.\n   - **Check 2: Prose Inline Path Integrity**:\n     - Strip fenced markdown codeblocks (` ```...``` `).\n     - Extract all backtick inline code spans resembling paths (` `src\u002Fservices\u002Ftasks.py` `).\n     - Assert that every referenced path exists on disk (`Path.exists()`).\n   - **Check 3: Script Command Validity**:\n     - Extract all mentioned script invocations (e.g., `npm test`, `pytest`, `ruff check`).\n     - Verify against declared scripts in `package.json`, `pyproject.toml`, or `Makefile`.\n   - **Check 4: Generator Sync Drift**:\n     - Re-run `harness context sync --dry-run` or `sync_context.py --dry-run`.\n     - Fail if on-disk vendor files differ by even a single character from canonical generation.\n2. **Wire into CI Pipelines & Git Hooks**:\n   - Add context linting to GitHub Actions \u002F CI workflow:\n     ```yaml\n     - name: Verify Codebase Context Files\n       run: harness context lint --root .\n     ```\n   - Wire pre-commit hooks to block commits containing broken paths or drifted vendor files.\n\n\n> **Completion criterion**: Context linter passes with exit code 0, 0 broken paths, 0 missing scripts, and zero sync drift.\n\n---\n\n## 5. Empirical Benchmarking & Definition of Done Feedback\n\nGround agent execution with deterministic feedback loops and objective validation:\n\n1. **Formulate Verifiable Definitions of Done**:\n   - Never allow an agent to declare a task finished based on \"looks completed\".\n   - Mandate specific test\u002Flint execution commands in the root file.\n   - **Require Evidence**: Demand that the agent paste the actual terminal output rather than claiming success in prose.\n2. **Install PostToolUse Agent Hooks**:\n   - Where supported by the client harness (e.g. Claude Code `.claude\u002Fsettings.json`), install `PostToolUse` hooks on `Edit|Write` actions that silently run context linting:\n     ```json\n     {\n       \"hooks\": {\n         \"PostToolUse\": [\n           {\n             \"matcher\": \"Edit|Write\",\n             \"hooks\": [\n               {\n                 \"type\": \"command\",\n                 \"command\": \"python scripts\u002Fcontext_linter.py --silent\"\n               }\n             ]\n           }\n         ]\n       }\n     }\n     ```\n3. **Execute Empirical A\u002FB Benchmarks**:\n   - Validate context file ROI by testing a standardized feature implementation prompt under two conditions:\n     - **Branch A**: With full multi-layer context files active.\n     - **Branch B**: With context files deleted (baseline).\n   - Score the delta along 4 objective metrics:\n     1. Test pass rate on first attempt.\n     2. Number of human interventions\u002Fcorrections required.\n     3. Adherence to architectural boundaries and layering.\n     4. Zero unapproved third-party dependencies introduced.\n\n> **Completion criterion**: Definition of done enforced with pasted evidence, post-tool hooks wired, and empirical A\u002FB validation completed.\n\n---\n\n## The Three Foundational Pillars\n\n### 1. The Visual Brief Pillar\nEvery major context restructuring or repository onboarding session must synthesize the context topology into an interactive HTML visual brief in `%TEMP%` loading Tailwind CSS and Mermaid.js via CDN. The brief visualizes the 3 layers, token budgets, and linter status.\n\n### 2. The Mandatory Checkpoint Gate Pillar\nWhen re-architecting repository context files, partitioning rules, or configuring CI gates, author an `implementation_plan.md` artifact with `RequestFeedback: true`. The agent must **STOP and wait** for explicit developer review and approval before mutating workspace files.\n\n### 3. Explicit Anti-Patterns\nRigid behavioral boundaries must be maintained to prevent context rot, instruction pollution, and ungrounded agent drift.\n\n---\n\n## Anti-Patterns\n\n- **Kitchen Sink Bloat** — Inlining generic tutorials, language default lessons, or team history into root files, exhausting model attention budgets.\n- **Silent Context Rot** — Allowing stale backtick paths and deleted scripts to linger because no CI check or pre-commit hook verifies them.\n- **Vendor Fork Divergence** — Maintaining separate hand-edited copies of CLAUDE.md, AGENTS.md, and Copilot instructions that contradict each other.\n- **Vague Encouragement Trap** — Stating ungrounded platitudes (\"write maintainable code\") instead of concrete boundaries, patterns, and worked examples.\n- **Inlined Monolith Sprawl** — Inlining multi-page architecture guides and schemas into root instructions instead of pointing to on-demand docs.\n- **Unverifiable Looks-Done Exit** — Allowing agents to finish tasks without running concrete test\u002Flint commands and pasting verifiable CLI output evidence.\n",{"name":5,"description":7},"main","eff3b00e36d1d233ad4a4389605c82fbfa0326d3",".agents\u002Fskills\u002Fcodebase-context-architect","https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness","https:\u002F\u002Fgithub.com\u002FSpectreDeath","I'm here to manage stress so I wouldn't expect earth shattering break throughs I'm just an enthusiast that likes making stuff. ","2026-11-15 05:32:58",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add SpectreDeath\u002FBrain-Harness --skill codebase-context-architect",{"tool":37,"command":38},"bunx","bunx skills add SpectreDeath\u002FBrain-Harness --skill codebase-context-architect",{"tool":40,"command":41},"pnpm","pnpm dlx skills add SpectreDeath\u002FBrain-Harness --skill codebase-context-architect","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002FSKILL.md",true,[45,51,55,59,63,69,73],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"CARD.md","REFERENCE","参考文档",3623,"https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002FCARD.md",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"config.default.yaml",1841,"https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002Fconfig.default.yaml",{"path":56,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":57,"sourceHtmlUrl":58},"references\u002Fcontext-linter-guide.md",2841,"https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002Freferences\u002Fcontext-linter-guide.md",{"path":60,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":61,"sourceHtmlUrl":62},"references\u002Fthree-layers-architecture.md",5252,"https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002Freferences\u002Fthree-layers-architecture.md",{"path":64,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":67,"sourceHtmlUrl":68},"scripts\u002Fcodebase_context_engine.py","SCRIPT","脚本",1559,"https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002Fscripts\u002Fcodebase_context_engine.py",{"path":70,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":71,"sourceHtmlUrl":72},"scripts\u002Fcontext_linter.py",2626,"https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002Fscripts\u002Fcontext_linter.py",{"path":74,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":75,"sourceHtmlUrl":76},"scripts\u002Fsync_context.py",2435,"https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fcodebase-context-architect\u002Fscripts\u002Fsync_context.py",[78,85,92,99,106,113],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":19},"01M7E0W6FR68J2N9CXCZTAQJFX","ui-behavior-primitive-architect","Architect, audit, and verify zero-dependency headless UI behavior primitives, WebMCP agent bridges, WAI-ARIA\u002FSSR\u002FRTL tri-modal checks, and framework expert routing. Do not use for static CSS theme styling or backend database migrations.","https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fui-behavior-primitive-architect\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":19},"01M7E0VZ7JQNW34G271DENFGYP","trustgraph-ontology-governor","Govern knowledge graphs, SHACL-AF triple policies, OntologyRAG subclass validation, PROV-O dual named graphs, and Laya entropy decision scoring. Do not use for relational SQL queries or generic web scraping.","https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Ftrustgraph-ontology-governor\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":19},"01M7E0VQ6Z7R45Q8QAZ4MT39EK","shim-harness-governor","Govern universal CLI agent loops, 4-tier tool permissions, ancestor-symlink workspace sandboxes, deterministic-first command routing, collision-safe MCP namespacing, and layered skill loading. Do not use for static UI styling or external SQL analytics.","https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fshim-harness-governor\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":19},"01M7E0VC18NMP1QBSRJ3TPQ7MB","graphiti-zep-memory-governor","Govern bi-temporal knowledge graphs with Graphiti future-edge clamping, MinHash\u002FLSH entity dedup, RRF\u002FMMR reranking, and Zep typed ontologies. Do not use for static SQL joins or generic web crawling.","https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fgraphiti-zep-memory-governor\u002FSKILL.md",[],[],{"skillNo":107,"name":108,"nameZh":6,"description":109,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":110,"categories":111,"keywords":112,"publishedAt":19},"01M7E0V12ABE1RW3ZRCST8VH5Z","agora-mcp-workbench-architect","Architect, audit, and orchestrate Agora Workbench MCP servers, loopback sidecars, unified state graphs, gateway\u002Fdispatcher connectors, and policy-enforced data lakes. Do not use for static frontend CSS styling or non-MCP desktop GUI automation.","https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fagora-mcp-workbench-architect\u002FSKILL.md",[],[],{"skillNo":114,"name":115,"nameZh":6,"description":116,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":117,"categories":118,"keywords":119,"publishedAt":19},"01M7BFYG6QT3484VC57TZ19V7P","skill-eval-assurance-runner","Evaluate and assure agent skills using small-N Wilson confidence intervals, unbiased pass^k reliability, deterministic trajectory graders, preflight boundary honesty, and sealed digest verification. Do not use for subjective prose review or unconstrained production deployment.","https:\u002F\u002Fgithub.com\u002FSpectreDeath\u002FBrain-Harness\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fskill-eval-assurance-runner\u002FSKILL.md",[],[]]