Reconcile mission objectives and steers against diffs and test results.
npx skills add https://github.com/MathieuDoyon/mission --skill mission-review
Please help me install this Agent Skill. Skill: mission-review Source: https://github.com/MathieuDoyon/mission/tree/main/.claude/skills/mission-review Command: npx skills add https://github.com/MathieuDoyon/mission --skill mission-review
It prevents a coding mission from being marked complete when the changes only partially satisfy the objective, missing acceptance criteria, or failing broader verification.
Use the mission-review sub-skill to produce a mission-report that verdicts complete, continue, or blocked after comparing the objective and steered requirements against the current diff and verification results.
High-intent search queries and answers about installing and using this skill.
Mission completion verification requires reconciling the original objective against the current diff and running broad test verification. This process inspects for scope creep, missing required pieces, and leftover TODOs to ensure changes fully satisfy acceptance criteria.
A quality gate for orchestrated coding evaluates whether a mission is truly done by comparing the diff against active steers and running full test commands. It produces a structured verdict indicating whether to proceed, continue, or stop due to missing context.
Automated diff analysis re-reads the objective and active steers, then inspects the diff for scope creep and missing required pieces. It runs the project's full verification command and treats any test failures as evidence that work must continue.
Run a mission review checkpoint when completion is plausible, after meaningful changes land in the diff, or optionally after related delegated TDD work finishes. This ensures the objective and steered requirements are met before finalizing.
Test verification during mission review runs the project's relevant full test command alongside static checks when applicable. Verification outcomes are captured and any failures are treated as evidence that the coding mission must continue rather than complete.
When a mission review detects missing context or failed verification, it emits a blocked verdict in the mission-report. The orchestrator is then directed to stop and follow the recommended next step rather than marking the mission complete.
hotyanovichaFetch pull request details and diffs to review code changes.
ResearAICreates canonical records for quest decisions with verdicts, actions, and evidence paths.
AutumnsGroveOrchestrate the TDD red-green-refactor cycle with isolated sub-agents.
dbc-oduffyRun a pre-merge weekly code-review gate with parallel Sonnet reviewers and security, dependency, and test evidence workers.
