Tribunal

Plugin tribunal from marketplace xmartlabs in xmartlabs/agent-skills. Commands read from the repo's manifest — today.

Skills (1)

• review-pr

  Use when reviewing an open pull request in a repo that has a review profile — anyone's, including your own — e.g. \"review PR #1851\" or /tribunal:review-pr <number>. Not for reviewing your own uncommitted working diff (dispatch the `tribunal:code-review-expert`, `tribunal:fe-code-review-expert` or

Subagents (9)

• code-review-expert

  Reviews recently written or modified BACKEND code (Node, Express and Prisma: controllers, services, schema and migrations, background jobs) before committing / opening a PR. Its FE sibling `fe-code-review-expert` covers frontend repos; pick by which repo the changed files live in, never by which one

• e2e-code-review-expert

  Reviews recently written or modified browser END-TO-END tests (Playwright and TypeScript: specs, fixtures, page objects, helpers, playwright.config) before committing / opening a PR. Its siblings `code-review-expert` (backend) and `fe-code-review-expert` (frontend) cover application repos; pick by w

• fe-code-review-expert

  Reviews recently written or modified FRONTEND code (React components, hooks, pages, styles, API integration) before committing / opening a PR. Its BE sibling `code-review-expert` covers backend repos; pick by which repo the changed files live in, never by which one the task mentions. Invoke explicit

• finding-verifier

  Decides whether ONE previously-raised finding is fixed at a given commit. Not a reviewer — it answers a single closed question about a claim someone else already made, and it must quote code to answer it. Used by the peer-review skill's incremental re-review.

• py-code-review-expert

  Reviews recently written or modified PYTHON code (services, APIs, workers, data jobs, migrations, tests; Django, FastAPI, Flask, SQLAlchemy and plain scripts) before committing / opening a PR. Its siblings `code-review-expert` (Node backend), `fe-code-review-expert` (frontend) and `e2e-code-review-e

• tribunal-mechanic

  Mechanical step runner for tribunal's PR review (the Rust engine): every dispatch tagged MECHANICAL, such as fetch-pr, context-bookkeeping, the worktree setup and teardown, the external-head relays, the anchor resolvers and reconcile-audit-persist. It runs a shell command the script already decided

• tribunal-refuter

  Adversarial refuter for one panel slot in tribunal's PR review (the Rust engine). Not a general code reviewer: its only job is to try to kill ONE already-raised finding by finding the input, code path, or schema fact that proves it is NOT real.

• tribunal-scribe

  Text step runner for tribunal's PR review (the Rust engine): the dispatches whose whole input is already in the prompt, namely `synthesize` (the report a person reads), `dedup-candidates` (clustering findings that make the same claim) and `panel-ceiling-tiebreak` (ranking findings tied at the panel

• tribunal-strengthener

  Refine step of tribunal's PR review (the Rust engine): strengthens ONE finding whose refutation panel came close to killing it although its core looked real, by reading the code at the PR head and tightening the claim, the repro and the severity, or by saying it cannot be made stronger. It raises no

Similar plugins