Review Agent

Code review specialist that evaluates pull requests for correctness, security, intent alignment, style, and documentation currency.

How the agent works

The review agent is triggered when a PR is opened or updated. It follows the same pre-script / sandbox / post-script pipeline as the other agents.

1. Pre-script validates inputs and fetches PR metadata.
2. Sandbox — the agent runs the pr-review orchestrator skill. The orchestrator triages the change, then dispatches specialized sub-agents in parallel — each covering a distinct review dimension (correctness, security, intent & coherence, style & conventions, docs currency, and optionally cross-repo contracts). Sub-agents run concurrently and return structured findings. The orchestrator collects, deduplicates, and synthesizes findings across dimensions, runs PR-level checks (scope authorization, protected paths), and produces a structured JSON review result. The agent cannot push files, edit code, or push — it is strictly read-only.
3. Validation loop — the output is checked against a schema, with up to 2 retry iterations if the output is malformed.
4. Post-script posts the review on the PR.

If a prior review exists (e.g., re-review after fixes), it is injected into the sandbox so the agent can assess whether previous findings were addressed.

How it helps

• Every PR gets a thorough review within minutes, regardless of team availability.
• Reviews cover security, correctness, intent & coherence, style, and docs currency — dimensions humans sometimes skip under time pressure.
• The structured output format makes it easy to see what was flagged and why.

Commands

Command	Where	Effect
/fs-review	PR comment	Triggers a review on the PR (per-repo installs only; standalone issues are ignored)

Requires triage-level repository permission or higher (triage, write, maintain, or admin). Mutation stages such as /fs-fix still require write or higher.

The /fs-review command does not accept arguments. The review agent also runs automatically when a PR is opened, synchronized (new commits pushed), or moved out of draft by a user with triage-level repository permission or higher — subject to the skips below. On GitLab, automatic review fires when the cron poller sees an MR whose created_at is newer than the watermark (up to one poll interval of delay). Native merge_request_event dispatch was removed; all GitLab events route through the poller. Push-to-open-MR (GitHub synchronize) is not auto-detected; comment /fs-review to re-review after new commits.

Automatic skips

The review agent does not run automatically — though /fs-review always works — when:

• The PR is a draft. Opening or pushing to a draft does not trigger a review; it runs once the PR is marked ready for review. (Applying the ready-for-review label to a draft lifts the draft skip, but the label is an automatic trigger like any other and is still subject to the fullsend-no-review and documentation-prose skips below. Only /fs-review is an unconditional override.)
• The PR carries the fullsend-no-review label. See Control labels below.
• The diff is documentation prose only. A PR whose changed files are all markdown under docs/guides/, docs/problems/, docs/agents/ or docs/glossary.md — and whose pages carry no executable markup — is skipped with a notice in the job summary. Prose is an allowlist: markdown anywhere else under docs/ (ADRs, normative/, contributing/, reference/, cli/, architecture.md, .vitepress/, …), markdown elsewhere in the tree (skills/*/SKILL.md, AGENTS.md, CLAUDE.md are executable agent instruction) and lockfiles (a lockfile-only diff can repoint a dependency) are all still reviewed. Because VitePress compiles every page under docs/ into a Vue component, each allowlisted page is also read at the PR head and keeps its review if it contains a <script> or <style> block, a head: frontmatter key, {{ }} interpolation, or a bound attribute or directive on raw HTML outside code. A file renamed into docs/ is judged on the path it came from, and a page or file listing that could not be read, or a listing that was truncated, never skips.

A push that is skipped for any of these reasons still clears ready-for-merge and ready-for-review, exactly as a review round would at start — the labels never describe commits nobody reviewed.

See ADR 0096 for the rationale.

Control labels

Most of these labels are applied by the review post-script based on the review outcome; fullsend-no-review is the exception — see its row below.

Label	Meaning
ready-for-review	Workflow state marker on the PR. Applied by the code agent post-script after pushing. In per-repo installs, triggers review when applied to a PR.
ready-for-merge	The review agent approved the PR. No blocking findings.
requires-manual-review	The review agent found issues that require human judgment — it could not confidently approve or reject.
rejected	The review agent rejected the PR and the post-script closed it.
fullsend-no-review	Prevents automatic (bot-triggered) review runs on this PR. Mirrors the fix agent's fullsend-no-fix label. Explicit /fs-review commands are unaffected. Currently applied manually (no /fs-review-stop command yet — see ADR 0096).

When the review agent requests changes (without rejecting), no outcome label is applied — the pull_request_review event triggers the fix agent directly.

Stale outcome labels from prior review runs are removed before the new one is applied.

The issue-labels skill may also apply contextual labels (e.g., area/api, priority/high) but these are informational -- they do not control agent behavior.

Configuration and extension

Skill: issue-labels

The review agent includes the issue-labels skill to discover your repo's labels and apply them to PRs during review. This is the same skill used by the triage agent -- overloading it affects both agents.

To overload the built-in skill, create your own issue-labels skill in .agents/skills/issue-labels/SKILL.md and symlink .claude/skills to .agents/skills so it's discoverable by both fullsend and local agent tooling. You can also overload it at the org level using config-driven agent registration -- see Bring Your Own Agent.

See Configuring with AGENTS.md and Configuring with Skills.

Variables

Variable	Description	Default	Valid values
REVIEW_FINDING_SEVERITY_THRESHOLD	Minimum severity for findings to include in the review. Findings below this level are omitted from both the narrative body and the posted inline comments.	low	info, low, medium, high, critical

Set this in the harness's env.sandbox (the upstream default lives in harness/review.yaml). To override per repo or org, use base: composition rather than the CI workflow env: block — workflow env: is reserved for infrastructure plumbing (see Architecture for details on harness composition and workflow-env conventions). The post-script reads the value from the runner environment directly — no separate configuration is needed.

The review agent omits findings below the threshold from its output. The post-script also filters the structured findings array as defense-in-depth. When filtering removes all findings from a request-changes or reject verdict, the post-script downgrades the verdict to comment (applying the requires-manual-review label).

Source

fullsend-ai/agents — harness/review.yaml