Review atomic work beads against requirements and ticket scope. Only this skill can close accepted work.

Install

mkdir -p .claude/skills/beo-review && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/11799" && unzip -o skill.zip -d .claude/skills/beo-review && rm skill.zip

Installs to .claude/skills/beo-review

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Review one executed atomic BEO bead against its self-contained description, TICKET.yaml scope, evidence, verification, and done criteria. Emits one route; only this skill may close accepted work.
195 charsno explicit “when” trigger
Advanced

Key capabilities

  • Review executed BEO beads against self-contained descriptions
  • Verify implementation against TICKET.json scope
  • Check if done criteria are covered by evidence
  • Confirm verification results support done criteria
  • Audit changed files against approved scope and generated outputs
  • Emit exactly one review route based on findings

How it works

The skill performs a fresh read of the bead's context, then applies a rubric to evaluate scope, intent, done criteria, and verification. It audits changed files and emits a single review route based on the findings.

Inputs & outputs

You give it
An executed atomic BEO bead, its TICKET.json, state, and runtime events
You get back
Exactly one review route (e.g., verdict_accept, repair_same_scope, user_review_needed)

When to use beo-review

  • Verifying bead implementation against scope
  • Checking completion criteria for tasks
  • Closing accepted work items
  • Ensuring worktree isolation compliance

About this skill

beo-review

Read

  • beo-reference -> references/default-reads.md
  • .beads/artifacts/<parent-issue-id>/PLAN.md only when review evidence or an explicit inconsistency requires parent-boundary audit; child descriptions should be self-contained
  • .beads/beo-reservations.jsonl and beo-reference -> registry/reservation-schema.json before any route that may release an existing reservation
  • beo-reference -> scripts/beo_worktree.py when the bead has worktree_isolation: true (for merge or cleanup)
  • beo-reference -> registry/state.schema.json when state update ownership or fields are unclear
  • beo-reference -> scripts/beo_check.py before verdict_accept (run --check review-entry --issue <id> to mechanically enforce verify/behaviour_gate passage and strict cross_check)

Do

  1. Fresh-read br, ticket, state, runtime events when present, phase-relevant registries named above, and any referenced evidence. If the atomic bead was decomposed from an epic/feature, review against the self-contained child bead description; read the parent PLAN.md only when the child description is ambiguous, review evidence explicitly references it, or a suspected parent-boundary conflict cannot be resolved from the child bead and TICKET.json.
  2. Apply the review rubric:
    • Scope: every changed file is allowed by TICKET.json or declared generated outputs.
    • Intent: the implementation satisfies the self-contained child bead description.
    • Done criteria: each criterion is covered by evidence or explicitly marked not covered.
    • Verification: recorded command results (including any declared scope.behaviour_gate) actually support the done criteria, not merely command execution. If scope.behaviour_gate is declared, its result must be present and passing before verdict_accept.
    • Regression surface: obvious adjacent behavior affected by the touched files is considered.
    • Repair boundary: same-scope repair is allowed only when file set, generated outputs, done criteria, verification, mode, risk, and Human Gates remain unchanged. Done when: each of the 6 rubric items evaluated and recorded as evidence-backed pass/fail.
  3. Audit changed files against approved scope and generated outputs; confirm verification results cover scope.verify.commands and done_criteria; record compact done-criteria coverage. Only emit verdict_accept when scope, intent, done criteria coverage, and verification evidence all support acceptance, and — for strict mode — a review.cross_check second-reviewer signal with verdict: agree is recorded (kernel §15; if no second reviewer is available, route user_review_needed). Mechanically confirm via beo_check.py --check review-entry --issue <id> --root <root> before verdict_accept; it enforces verify and behaviour_gate passage and, for strict, the cross_check signal. If evidence is missing but work may be correct, route repair or user decision; do not accept on trust. Done when: every changed file in approved scope or generated outputs; verification covers both scope.verify.commands and done_criteria; coverage recorded; verdict_accept only when all 4 conditions (scope, intent, coverage, evidence) hold.
  4. Record findings with severity, category, message, evidence refs, and recommended route; the final route must be derivable from findings. For user_review_needed, the route may be derived from review.route_condition_id, blocking findings with recommended_route: none, and the Beads decision envelope. Emit exactly one review route.
  5. For root_cause_diagnosis_needed, set the route condition, leave review.verdict null, and append a handoff runtime event before routing to beo-debug. Use repair_same_scope only when approved files, generated outputs, done criteria, verification, mode, risk, and Human Gates remain unchanged; otherwise use repair_rescope.
  6. For beads with worktree_isolation: true:
    • On verdict_accept: merge the worktree branch via beo-reference -> scripts/beo_worktree.py merge --issue <issue-id>, then cleanup via beo-reference -> scripts/beo_worktree.py cleanup --issue <issue-id> --reason accepted.
    • On repair_same_scope or repair_rescope: cleanup without merge via beo-reference -> scripts/beo_worktree.py cleanup --issue <issue-id> --reason repair.
    • On cannot_deliver or abandoned: cleanup via beo-reference -> scripts/beo_worktree.py cleanup --issue <issue-id> --reason <route>.
  7. If during review a BEO harness improvement is identified:
    • Write .beads/artifacts/<issue-id>/harness-proposal.json following beo-reference -> registry/harness-proposal.schema.json.
    • Emit harness_change_needed -> beo-author. The review is paused; beo-author returns to caller after resolution.
    • On return from beo-author, re-read state and continue review.
  8. Close with br only on verdict_accept; otherwise leave the issue open for repair or user action. When emitting user_review_needed, set review.route_condition_id to user_review_needed, leave review.verdict null, record blocking findings with existing categories and recommended_route: none when no enum route applies, and follow the user_review_needed handoff format in beo-reference -> references/user-handoff.md. Do not write a user_stop runtime event for review user decisions.

Write

  • state.json phase and review fields only
  • .beads/artifacts/<issue-id>/harness-proposal.json when proposing a harness change
  • Beads comments/labels for the final route when needed, including compact user_review_needed route comments and labels only when an existing BEO label represents the state
  • Reservation release on verdict_accept, cannot_deliver, abandoned, and repair_rescope only for strict-mode active reservations or when a reservation exists
  • runtime-events.jsonl for non-normal review events, including handoff before root_cause_diagnosis_needed
  • Optional learning_candidate only after a final route and only when high-signal

Emit

  • verdict_accept -> close accepted work
  • repair_same_scope -> beo-validate
  • repair_rescope -> beo-plan
  • cannot_deliver -> user
  • abandoned -> user
  • root_cause_diagnosis_needed -> beo-debug
  • harness_change_needed -> beo-author
  • user_review_needed -> user

Non-normal runtime-events.jsonl events (advisory, optional): verification_run (when beo_verify.py is invoked during review), intervention (when external human, CI, or reviewer input is recorded).

Never

  • See beo-reference -> registry/phase-contracts.json must_not[]; audit C8 enforces drift.
  • Do not mutate product files.
  • Do not grant PASS_EXECUTE.
  • Do not close non-accepted work.
  • Do not emit more than one review route.

When not to use it

  • When the task is not reviewing an executed atomic BEO bead
  • When the skill is asked to mutate product files
  • When the skill is asked to close non-accepted work

Limitations

  • May only close accepted work
  • Must not mutate product files
  • Must not grant PASS_EXECUTE

How it compares

This skill provides a strict, rubric-based review process for BEO beads, ensuring all criteria are met and evidence is present before work is accepted, which is more rigorous than a general code review.

Compared to similar skills

beo-review side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
beo-review (this skill)01moNo flagsAdvanced
resolve-conflicts818moReviewIntermediate
claude-automation-recommender472moReviewBeginner
codex-skill125moReviewAdvanced

Try saying

Example prompts that trigger this skill in your AI assistant.

You might also like

resolve-conflicts

antinomyhq

Use this skill immediately when the user mentions merge conflicts that need to be resolved. Do not attempt to resolve conflicts directly - invoke this skill first. This skill specializes in providing a structured framework for merging imports, tests, lock files (regeneration), configuration files, and handling deleted-but-modified files with backup and analysis.

81334

claude-automation-recommender

anthropics

Analyze a codebase and recommend Claude Code automations (hooks, subagents, skills, plugins, MCP servers). Use when user asks for automation recommendations, wants to optimize their Claude Code setup, mentions improving Claude Code workflows, asks how to first set up Claude Code for a project, or wants to know what Claude Code features they should use.

47140

codex-skill

feiskyer

Use when user asks to leverage codex, gpt-5, or gpt-5.1 to implement something (usually implement a plan or feature designed by Claude). Provides non-interactive automation mode for hands-off task execution without approval prompts.

12110

git-advanced-workflows

wshobson

Master advanced Git workflows including rebasing, cherry-picking, bisect, worktrees, and reflog to maintain clean history and recover from any situation. Use when managing complex Git histories, collaborating on feature branches, or troubleshooting repository issues.

1197

subagent-driven-development

davila7

Use when executing implementation plans with independent tasks in the current session

1493

validate-openapi-specs

epieczko

Validates and registers hook manifest files (YAML) in the Hook Registry for versioned hook management.

594

Search skills

Search the agent skills registry