ai-coding-tooling

Extension for the mermaid-studio skill. This file MUST be read together with the parent mermaid-studio SKILL.md. The parent governs Mermaid diagram creation, validation, and rendering. This extension adds a required substitute renderer for multi-row C4 diagrams.

Perform comprehensive code reviews on implementation code. Reviews local workspace changes by default, or a GitHub PR when a PR number is provided. Covers architecture, performance, code quality, API design, and security. Also performs full-codebase Performance Audits (P0–P3 findings report) when triggered by performance audit phrases. Technology agnostic — adapts to the project's stack using context files. Use when the user says "review my code", "code review", "check my code", "review my changes", "review this PR", or "review PR #123". Do NOT use for reviewing test files — use the tests-code-review skill for that.

Extension for the tlc-spec-driven skill. MUST be read together with the parent tlc-spec-driven SKILL.md. The parent governs spec-driven planning. This extension (1) declares how reference-level overlays load, (2) prioritizes planning before implementation for non-trivial work, (3) names feature artifacts `<TASK-ID>-<slug>`, (4) augments coding-principles with software-design/observability/stack-style references and routes security to the security-best-practices skill and UI work to web-design-guidelines, and (5) requires a persisted per-feature commit log (`commits.md`) of every atomic commit hash traceable to that feature.

Review test code quality, coverage patterns, and maintainability. Ensures tests are clear, independent, and provide meaningful coverage. Technology agnostic — adapts to the project's stack using context files. Use when the user says "review tests", "test code review", "check tests", "review test coverage", "review my tests", "review tests on PR #123", or "check tests PR #42". Do NOT use for writing new tests — use the tests skill for that. Do NOT use for reviewing implementation code — use the code-review skill.

Checks whether a task prompt follows the anatomy-of-a-prompt guidelines (9 layers: Task, Context Files, Reference, Success Brief, Rules, Conversation, Plan, Alignment, Push Intent). Auto-invoked before responding to any new task. Outputs a quality report — confirmed or missing layers — then proceeds without blocking. Never invoke for follow-ups, continuations, or responses to clarifying questions.

Creates, updates, and incrementally syncs the project context documentation that agents load at session start. Three modes. Full mode deep-scans the codebase (brownfield mapping) and writes nine context files to docs/codebase/ — PROJECT.md (overview, vision, goals), STACK.md, STRUCTURE.md, ARCHITECTURE.md, CONVENTIONS.md, INTEGRATIONS.md, TESTING.md, CONCERNS.md, and PIPELINE.md. Incremental mode inspects the git workspace and syncs only what changed — inline API docs in source files, root context files (README.md, CLAUDE.md, AGENTS.md), and the context files in docs/codebase/ — and detects new packages. Package mode generates a scoped CLAUDE.md for an individual package/module. Use when the user says "evaluate architecture", "map codebase", "analyze existing code", "document current architecture", "update architecture docs", "refresh project context", "onboard project", "create project docs", "update project docs", "update docs", "document my changes", "sync documentation", "document recent changes", "evalua

Write and maintain tests for any project. Covers unit tests, integration tests, code coverage analysis, and TDD (test-driven development). Technology agnostic — adapts to the project's stack using context files. Use when the user says "write tests", "add tests", "missing tests", "test coverage", "unit test", "integration test", "TDD", "test-driven", "red-green-refactor", "write test first", or "test-first". Do NOT use for reviewing existing test quality — use tests-code-review for that. Do NOT use just to run tests.

Capture project-specific, opinionated insights about a technology into skill reference files. Skills load these files at invocation time if they exist. Use when the user says "save this insight", "add this to the [tech] reference", "remember this [tech] pattern", "capture this convention", or "log this for [tech]". Do NOT use to generate generic best practices from public documentation.

Create, update, plan, and resolve technical debt reports as a permanent numbered ledger. Reporting a debt adds a TD-XX row to docs/TECH_DEBTS.md (a documentation ledger of known debts — it is NOT auto-loaded into context at session start); the per-debt solution plan is written to docs/tech-debts/TD-XX.md only when you plan the fix. Rows are never deleted — only Status and Solved change. Use when the user says "tech debt report", "report tech debt", "document tech debt", "add tech debt", "update tech debt", "plan tech debt", "resolve tech debt", or "mark tech debt as resolved". Do NOT use for fixing tech debt, code reviews, or refactoring.

Extension for the skill-architect skill. This file MUST be read together with the parent skill-architect SKILL.md. Adds two capabilities: (1) guardrail design guidance injected into the existing workflow phases, and (2) the extended/ pattern for modifying globally installed skills without touching the source.

Token-efficiency extension for the docs-writer skill. This file MUST be read together with the parent docs-writer SKILL.md. Adds output rules to keep AI-agent-facing .md files lean.