Mission Verify

mission-verify

Requires: the factory-mission skill's shared executor contract at .agents/skills/factory-mission/references/executor.md (installed path); install with adlc-cli skills add tikalk/adlc-team-skills --skill factory-mission.

In a local session — you answer the gates

When ADLC_RUN_ID is not set, a person is in this session, and nothing else will answer a gate. The agentic-container runtime that relays gates to Slack exists only in a runner pod.

1. A paused run is your turn, not the end of the mission. When adlc-cli workflow run or resume exits with the run paused at a gate (run_paused, status: paused in adlc-cli workflow state show <run_id>), ask the person the gate's question and options. Then run adlc-cli workflow resume <run_id> --input <verdict_input>=<answer> and keep going. Never end your turn on a paused run that nobody was asked about.
2. Every gate that approves MRs is followed by a step that opens them. A gate whose message promises MRs/PRs is followed by a publish step: run-branch.sh and open-mr.sh for each repository the change touches, writing mr_url (mr_urls for more than one). Only then comes the write-back step, so the completion comment links real MRs. A gate with nothing after it to act on its answer is a plan defect; fix the plan before you start the run.
3. Before ending your turn, the run is completed, failed or aborted, or it is paused at a gate the person has been asked about.

What this skill does

mission-verify owns the verify phase of the mission SDD family: it judges the built work against the brief rather than the implementer's claims, emits severity-ranked findings, and returns one converge verdict. It works under the shared executor contract (.agents/skills/factory-mission/references/executor.md): the verify step reads the test and code artifacts and emits findings, and the converge step reads brief.md and emits a decision — the judge is independent, so it checks the brief and non-goals rather than the verifier's own findings. It stops at the verdict — REVIEW and SHIP belong to factory-review and factory-deploy, which this skill never pre-empts.

Default lane

1. Read the artifact refs produced by the test and code steps.
2. Judge against brief.md, not the implementer's claims — re-run the suite, inspect the diff, and confirm the non-goals are untouched. The converge judge works from the brief and non-goals, never from the verifier's own findings.
3. Emit severity-ranked findings as Critical, Important or Minor, each tied to evidence in the diff, the suite output or an untouched non-goal.
4. Return one converge verdict: DONE when the criteria are met and the quality threshold is reached, CONTINUE when in-scope gaps remain, SPEC_CORRECTION_NEEDED when a criterion is unmeasurable or the scope conflicts.
5. Route SPEC_CORRECTION_NEEDED to the owning clarify gate, never to code; return CONTINUE to the implementation phase with the findings attached.
6. Never approve or merge; approval and merge stay with the human and the deploy phase.

Step contract

Gates

This phase's gate is the converge verdict — a decision gate, not a mechanical red/green gate (the mechanical gates bind the build steps in mission-implement).

• DONE — the acceptance criteria are met and the quality threshold is reached.
• CONTINUE — in-scope gaps remain; the findings return to implementation and the circuit breaker applies.
• SPEC_CORRECTION_NEEDED — a criterion is unmeasurable or the scope conflicts; the verdict routes to the owning clarify gate, never to code.

Boundaries

• No worktree: isolation belongs to the orchestrator; this skill works in the tree it is handed.
• No lease: never acquire, renew or release a run lease.
• No tracker writes: no comments, labels, status transitions, or MR/PR creation.
• No merge or approval: this skill never approves, merges or ships.
• Routes back: findings route to the phase that owns the gap; CONTINUE returns to implementation; SPEC_CORRECTION_NEEDED routes to the owning clarify gate, never to code.