AgentsKit Code Review
For agents

Code Review agent handoff

Ownership, change routes, checks, and security boundaries for the Code Review repository.

Code Review CLI handoff

Purpose

Provider-neutral, low-noise AI code review for local Git diffs, files/stdin, and GitHub pull requests. Bounded diff-first context packs run one multidimensional analysis; adversarial votes remove weak findings; reporters emit Markdown, GitHub reviews, and SARIF.

Ownership map

  • src/cli.ts: public flags, source selection, provider selection, exit policy.
  • src/campaign-reducer.ts and src/campaign-store.ts: deterministic lifecycle, atomic checkpoints, leases, replay, and resume.
  • src/campaign-preflight.ts: provider-free campaign discovery, eligibility, source planning, budget checks, and worktree gating.
  • src/batch-mode.ts: shared measured batching, including single-file context-pack partitioning (ADR-0006); preserve pack identity in artifacts, cache keys, and consolidation.
  • src/budget.ts: provider-free hierarchical budget compilation, reservations, and dimension-aware usage accounting.
  • src/review-stores.ts: separate self-hosted feedback and approved-knowledge stores with bounded schemas, atomic writes, retention, concurrent-write serialization, and the scoped AgentsKit Retriever.
  • src/review-feedback.ts: deterministic feedback reconciliation, resumable checkpoints, metrics, and inactive candidate-rule evidence.
  • src/campaign-runner.ts and scripts/review-campaign.mjs: bounded multi-PR execution, leased atomic checkpoints, resume, per-PR isolation, and deterministic terminal reporting.
  • src/quality-matrix.ts: run and campaign quality scoring, evidence separation, cost/time accounting, and baseline regression gates. evaluateCampaignQuality() must be used when aggregating campaign outcomes.
  • src/<provider>-adapter.ts: logged-in local CLI adapters.
  • src/provider-registry.ts: validated provider capabilities and conservative execution policy around AgentsKit adapters.
  • src/provider-execution.ts: typed provider failures, bounded retry delay, adaptive concurrency, and cancellation-aware queueing.
  • src/scm-contract.ts: provider-neutral change-request, publication, readiness, and merge contracts plus explicit capability checks.
  • src/github-scm-adapter.ts: GitHub REST/CLI binding for the common SCM contract; the only migrated path allowed to contain GitHub payloads or merge commands.
  • agents/code-review/: review pipeline, lenses, input normalization, reporters.
  • agents/code-review/risk.ts: provider-free context-pack risk signals and specialized-depth selection.
  • action.yml: composite GitHub Action contract.
  • .github/workflows/release.yml: Changesets version-pull-request workflow.
  • .github/workflows/publish.yml: version-PR-gated npm Trusted Publishing and GitHub Release workflow.
  • examples/: copy-ready Action workflows.
  • README.md and docs/OPERATIONS.md: public adoption and operations guidance.
  • ecosystem.json, llms.txt, and llms-full.txt: canonical product graph and machine-readable discovery/full-corpus surfaces.
  • test/: credential-free CLI, Action, and documentation contract proofs.

Boundaries

  • Depend on AgentsKit adapter/runtime/tool contracts; do not create a second model abstraction.
  • Preserve provider neutrality and advisory-by-default Action behavior.
  • Never expose provider keys in arguments, docs fixtures, logs, or PR output.
  • The Fumadocs site provides curated documentation only; the product runtime remains the CLI and GitHub Action. Do not add an embedded AgentsKit Chat runtime here.
  • The vendored review agent tracks the AgentsKit Registry source; keep divergences explicit.

Change routes

  • CLI flag/provider behavior: start at src/cli.ts, then update README, operations docs, and tests.
  • Cycle entry-point changes must pass the npm-style symlink subprocess test; an import must remain side-effect free.
  • Fatal batch/provider errors must abort peers and drain active work before writing a terminal result. CLI SIGTERM must reach the provider subprocess, not just terminate its parent.
  • Local CLI subprocess behavior: start at the matching src/<provider>-adapter.ts and add an offline fixture.
  • Review logic or noise reduction: start at agents/code-review/agent.ts, agents/code-review/sources.ts, and the relevant dimension prompt; preserve bounded diff-first packs, exact changed-line anchoring, pre-provider AgentsKit token checks, and one structured analysis per normal pack. Prove both survival and rejection behavior.
  • Campaign lifecycle or resume behavior: update the reducer/store together and run their focused crash/replay tests.
  • GitHub comments/SARIF: start at agents/code-review/reporters.ts and verify permissions/failure docs.
  • Action input: update action.yml, examples/pull-request.yml, README, and contract tests together.
  • Release automation: update .github/workflows/release.yml, .github/workflows/publish.yml, Changesets configuration, and the automated publishing section in docs/OPERATIONS.md together.

Verification

npm ci
npm run check
npm pack --dry-run

npm run check includes typecheck, build, an end-to-end offline stdin review, Action/documentation tests, Doc Bridge gates, and CLI help.

Ecosystem routes

Use llms.txt for discovery and llms-full.txt only when the complete public, operational, and agent-handoff context is required.

Human guide

Continuous improvement

Run evidence-based improvement cycles for Code Review quality.

On this page

Code Review CLI handoffPurposeOwnership mapBoundariesChange routesVerificationEcosystem routesHuman guide