TeaRAGs-MCP

Decide HOW to write new code by inspecting git signals on neighbors — pulls strategy and templates from proven, low-churn nearby implementations rather than guessing from training data. Triggers on "implement function X", "add method to class Y", "write a helper for Z", "напиши функцию", "добавь метод". NOT for discovery or exploration — use tea-rags:explore for that. This skill activates ONCE the agent is about to write code.

Agentic-only enrichment skill — surface battle-tested reference code from project as templates for generation / modification. Three-level locality cascade (target subdomain → domain → project), quality gate by overlay labels + project-wide proven rerank preset. Returns ranked list of reference chunks + locality annotation (L1 / L2 / L3 / none). Invoked by `tea-rags:data-driven-generation` Step 2 (TEMPLATE), `dinopowers:writing-plans` (per code-gen Task), `dinopowers:executing-plans` (per Task during execute). Skipped when no `positiveIds` / `positiveCode` and no `behaviorQuery` available.

Run automated TeaRAGs install wizard from scratch — detects environment, installs deps (Node.js, tea-rags, Ollama/ONNX, Qdrant), tunes performance, configures MCP server. Progress saves to ~/.tea-rags/setup-progress.json for resumable install. Triggers on "install tea-rags", "set up TeaRAGs MCP", "configure tea-rags from scratch", "поставить tea-rags в проект". NOT for tuning an existing install — use tune for that.

Agent-only. Raise test coverage above a failed pre-commit threshold: locate the

Brainstorm code change seeing risk, ownership, tech-debt signals from tea-rags first — creative exploration grounded in actual state of affected area. Triggers on "brainstorm X", "design feature", "refactor Y", "let's discuss", "давай обсудим", "как перестроить", "что может сломаться". NOT for trivial edits or stylistic questions with no code area to enrich. Wraps superpowers:brainstorming with tea-rags risk-signal enrichment step.

Execute written implementation plan whose Tasks edit code, per-Task SAFE/CAUTION/UNSAFE git-signal verdict before edit AND code-style cascade for code-generation Tasks (style from silo authors, strategy+template from proven neighbors). Triggers on "execute the plan", "start Task N", "выполни план", "начни задачу", "run the plan", "implement the plan steps". NOT for one-off edits without a written plan. Wraps superpowers:executing-plans with tea-rags git-signal verdicts and tea-rags:data-driven-generation cascade.

Finalize dev branch (merge/PR/cleanup) with tea-rags:risk-assessment over full branch diff — completion options weighed against risk zones across entire branch scope. Triggers on "finish the branch", "ready to merge", "wrap up the feature", "ветка готова", "доводим до merge", "branch ready", "PR time", "shipping the branch", "merge ready". NOT for mid-branch interim commits. Wraps superpowers:finishing-a-development-branch with tea-rags:risk-assessment over the branch diff.

Assess reviewer's suggested change (rename, refactor, move, extract) — query tea-rags imports/churn signals on affected target, measure blast radius before agreeing or pushing back. Triggers on "reviewer suggests", "should I rename", "ревьюер предлагает", "стоит ли переименовать", "code review feedback", "PR comment about", "review feedback says", "PR comment wants", "address the review", "сделай как просит ревьюер", "code review wants me to", "let's refactor this per review". NOT for vague stylistic nits without a named symbol. Wraps superpowers:receiving-code-review with tea-rags blast-radius signals.

Prepare code-review request bundle: per-file ownership, churn, connected

Debug concrete failure: first run tea-rags:bug-hunt for ranked suspect list

Write failing test BEFORE implementation, match project conventions from tea-rags search of existing test files (mocks, helpers, assertion style, fixture conventions) so new test fits local norms. Triggers on "write a test", "TDD", "напиши тест", "сначала тест", "failing test first", "RED phase", "implement feature X", "add test for Y", "fix bug Z". NOT for projects with no existing tests — fall back to superpowers:test-driven-development directly. Wraps superpowers:test-driven-development with tea-rags test-pattern search.

Before claiming work is complete, run tea-rags imports/churn signals on just-edited files to surface collateral-damage candidates (files whose dependents could break), then run verification commands so success claims are backed by evidence and blast-radius awareness. Triggers on "ready to commit", "all done", "should be fixed now", "готово", "проверь перед commit", "все работает". NOT for mid-work sanity checks without a completion claim. Wraps superpowers:verification-before-completion with a tea-rags collateral-damage scan.

Author implementation plan; Affected Files section enriched per-file imports, churn, ownership, bugFixRate, taskIds from tea-rags — task decomposition reflects real impact+risk. Triggers on "write a plan", "draft implementation plan", "напиши план", "распиши задачи по файлам", "plan the refactor". NOT for plans with no named files or code area to enrich. Wraps superpowers:writing-plans with per-file tea-rags signal enrichment.

Author/rewrite SKILL.md inside .claude-plugin/dinopowers/, enforce wrapper pattern (tea-rags enrichment step then chain into superpowers:*) + description style to win against parent skills. Triggers on "add dinopowers:X", "create a wrapper skill", "write a SKILL.md", "напиши скилл", "новый wrapper", "новый dinopowers скилл", "edit dinopowers wrapper", "rewrite dinopowers SKILL.md", "fix skill description", "improve a dinopowers skill", "tweak SKILL.md frontmatter". NOT for SKILL.md files outside .claude-plugin/dinopowers/. Wraps superpowers:writing-skills with dinopowers wrapper-pattern enforcement.

Benchmark this hardware and write optimal performance parameters

Pick a tea-rags rerank preset, build a `{custom: {...}}` weight set, or

Find source of concrete failure — rank historically buggy code (high bugFixRate + churn) against symptom. Triggers on "debug X", "why does Y fail", "test fails", "stack trace says Z", "падает", "почему не работает". NOT for code health scanning without a specific symptom — use risk-assessment for that.

Developer explore, understand, explain, investigate code — "how does X work", "show me the architecture of Y", "what does Z do", "where is X used", "find all X", "antipatterns in X", "best example of X". Also pre-generation investigation — "before I code/change/modify/refactor X", "what should I know before touching X", "research before coding", "context for changing X", "risks before refactoring X". NOT for active bugs (use bug-hunt), NOT for standalone code health scan without specific area (use risk-assessment)

Construct tea-rags filter beyond `pathPattern`. Invoke whenever agent

Force full re-index zero downtime. Builds new versioned collection in background while search continues on current. Alias switches atomically when done. Requires explicit user confirmation — NEVER invoke automatically.

Index/reindex codebase. First time — register alias + index in one command (index-codebase --name). Already indexed — incremental reindex (only changed files).

Help the user file a high-quality GitHub issue for a TeaRAGs bug — gather environment + diagnostic context, CHECK FOR EXISTING/KNOWN ISSUES FIRST, then produce a pre-filled issue URL (or `gh issue create` when available). Triggers on "report a bug", "file an issue", "open a GitHub issue about this error / the quarantine", "зарепорти баг в tea-rags", "это похоже на баг tea-rags". NOT for user-fixable config / setup errors (those hints are already actionable).

Assess project/domain health — scan risk zones across dimensions (bugs, hotspots, tech debt). Use when asked evaluate risks, find problematic areas, assess code health, identify zones needing attention — NOT for specific bug symptoms (use bug-hunt instead)

Agentic-only enrichment skill — surfaces DSL test chunks (chunkType "test"

Define new normalized 0-1 ranking signal computed from raw payload fields at rerank time, registered under trajectory provider (git or static). Triggers on "add freshness signal", "new churn metric", "score code by ownership", "новый сигнал для git churn". NOT for adding rerank preset combining existing signals — use add-rerank-preset for that.

Request/response shape exposed via public API to MCP tools, lives in api/public/dto. Triggers on "add DTO", "new request type for X tool", "add response shape", "новый DTO для tool'а". NOT for internal types — those live in contracts/, not public API surface.

Add chunker hook splitting a language's source files into searchable chunks (classes, functions, comments) in ingest pipeline. Triggers on "add Kotlin support", "improve TypeScript class chunking", "fix Ruby comment extraction", "add language X". NOT for tweaking generic chunker behavior — edit chunker base directly for language-agnostic changes.

Expose new tool to LLM clients via MCP protocol — schema, handler, App method, docs in one coordinated change. Triggers on "add MCP tool", "new endpoint that returns X", "expose Y as a tool", "новый MCP tool". NOT for modifying an existing tool's schema — just edit the tool file directly.

Add migration to existing pipeline (schema, snapshot, sparse) to upgrade persisted on-disk state. Triggers on "new Qdrant index", "backfill payload field", "snapshot format change", "rebuild sparse vectors", "новая миграция". NOT for code-only changes that don't touch persisted state on disk.

Named ranking strategy combining derived signals with weights + overlay mask, exposed as enum option to MCP tools. Triggers on "new tech-debt preset", "create hotspot detector", "security-audit ranking", "rank by ownership and churn". NOT for adding single signal — use add-derived-signal for that.

Analyze latest pipeline debug log for perf bottlenecks, timing issues, anomalies. Use when indexing slow, git blame too long, or pipeline behavior unexpected.

Eval-driven skill optimization. Use when skill needs improvement — wrong tool selection, missing use cases, verbose instructions, stale rules. Structured eval cycles with parallel subagents measure skill quality, iterate until 100% pass rate.

Index current tea-rags worktree as separate `tea-rags-worktree` project alias, verify end-to-end reindexing on worktree's own source. Auto-invoke when user asks to "test reindexing", "проверить реиндексацию", "test tea-rags self-reindex", "регрессия реиндексации", "test self-reindex on worktree", "smoke test enrichment on worktree", or any phrasing meaning "run live tea-rags self-test against current worktree's code". Use instead of indexing main `tea-rags` alias when exercising unreleased worktree code without disturbing main project's index.