
from repo-task-proof-loop715
A repo-local workflow skill that creates auditable task folders and runs a spec-freeze → build → evidence → verify → fix loop with fresh-session verification.
Provides a repeatable, auditable workflow for implementing and verifying non-trivial coding tasks inside a repository. The skill initializes .agent/tasks/<TASK_ID>/, installs project-scoped subagent templates, and enforces a spec-freeze, build, evidence, fresh verification, and minimal-fix loop that produces machine-readable verdicts and problems files.
Use for feature implementations, refactors, migrations, or bug fixes where you need repo-local proof artifacts, reproducible verification, and minimal trusted changes. Triggers: requests to implement significant tasks, run verifiers, or produce evidence-backed PASS/FAIL verdicts.
Designed for Codex/Copilot-style coding agents and Claude Code where repo-local subagents or child sessions can be used. Also useful for CI-integrated verifiers and automation that can run Python scripts in the repo.
This skill has not been reviewed by our automated audit pipeline yet.