metta:verify
Runs verification gates and spawns verifier subagents to check code against specs. Orchestrates compliance workflows from gate to shipment.
Install
mkdir -p .claude/skills/metta-verify && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/17196" && unzip -o skill.zip -d .claude/skills/metta-verify && rm skill.zipInstalls to .claude/skills/metta-verify
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.
Verify implementation against specKey capabilities
- →Run automated gates for verification
- →Spawn a `metta-verifier` subagent
- →Check Given/When/Then scenarios against tests and code
- →Write verification results to `summary.md`
- →Commit verification summary
- →Spawn a `metta-executor` to fix failures
How it works
The skill runs verification gates, spawns a `metta-verifier` subagent to check the implementation against the spec, writes results to a summary file, and commits it. If gates fail, it spawns a `metta-executor` to fix them.
Inputs & outputs
When to use metta:verify
- →Verifying changes against defined specs
- →Automating compliance checks
- →Managing verification gate failures
- →Finalizing and merging releases
About this skill
IMPORTANT: When using the Agent tool, use these metta agent types: metta-proposer (intent/spec), metta-researcher (research), metta-architect (design), metta-planner (tasks), metta-executor (implementation), metta-verifier (verification), metta-discovery (init). Do NOT use gsd-executor or general-purpose.
You are the orchestrator for verification. Spawn a verifier subagent.
Steps
Resolve {change_root} first: metta status --json --change <name> returns worktree — when non-null, that value is {change_root}; when null, the main checkout root is. Every path and git command below is anchored to {change_root} — never rely on the session cwd, which for a worktree-hosted change resolves to the wrong checkout.
metta verify --json --change <name>→ runs gates, returns results- Spawn a metta-verifier agent (subagent_type: "metta-verifier") with:
- The spec from
{change_root}/spec/changes/<change>/spec.md - The gate results
- Task: check each Given/When/Then scenario against tests and code
- Write results to
{change_root}/spec/changes/<change>/summary.md - Commit:
git -C "{change_root}" add "{change_root}/spec/changes/<change>/summary.md" && git -C "{change_root}" commit -m "docs(<change>): verification summary"
- The spec from
- Token recording is automatic — a SubagentStop hook records each subagent's harness-measured usage; do not run
metta tokens recordafter subagent returns. Only if the hook is unavailable, record manually:metta tokens record --task <artifact-or-task-id> --agent <subagent-type> --model <alias> --tokens <count> --change <name> --source prose. metta complete verification --json --change <name>- When all_complete: true, tell the user to run
/metta:shipto finalize and merge
If any gate fails or the verifier reports FAIL
Spawn a metta-executor to fix the failures, then re-verify from step 1. When the FAILing run's output was produced under a downgraded (non-inherit) model, before spawning the fix executor run metta model-escalation record --task <id> --from <resolved-model> --to inherit --trigger verify_fail --change <name>, then spawn the fix executor with the model parameter omitted (top-tier).
- If an executor or verifier STOP-reports a silent-write anomaly (Edit/Write success with no on-disk effect), escalate to the user with the report; never work around it via bash writes or orchestrator-performed writes.
After verification, the next step is always finalize (archive + spec merge), then ship (merge to main).
When not to use it
- →When not using `metta` agent types
- →When the change is not ready for verification
- →When the goal is not to finalize and merge
Limitations
- →Requires specific `metta` agent types
- →Verification results are written to `spec/changes/<change>/summary.md`
- →Relies on `metta` CLI commands
How it compares
This skill orchestrates a multi-step, agent-driven verification process with automated gates and specific subagent roles, which is more structured and automated than manual verification.
Compared to similar skills
metta:verify side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| metta:verify (this skill) | 0 | 4mo | No flags | Advanced |
| skill-creator | 128 | 4mo | Review | Advanced |
| dependency-upgrade | 26 | 5mo | Review | Intermediate |
| testing-workflow | 16 | 9mo | Review | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
skill-creator
anthropics
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
dependency-upgrade
wshobson
Manage major dependency version upgrades with compatibility analysis, staged rollout, and comprehensive testing. Use when upgrading framework versions, updating major dependencies, or managing breaking changes in libraries.
testing-workflow
amo-tech-ai
Comprehensive testing workflow for E2E, integration, and unit tests. Use when testing applications layer-by-layer, validating user journeys, or running test suites.
requesting-code-review
obra
Use when completing tasks, implementing major features, or before merging to verify work meets requirements
develop-ai-functions-example
vercel
Develop examples for AI SDK functions. Use when creating, running, or modifying examples under examples/ai-functions/src to validate provider support, demonstrate features, or create test fixtures.
finishing-a-development-branch
obra
Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup