From correctless
Generates one-page summary of Correctless feature workflow from specs, QA findings, verification reports, git logs, and audit trails. Use after /cdocs, mid-feature, or before PR.
npx claudepluginhub joshft/correctless --plugin correctlessThis skill is limited to using the following tools:
Generate a one-page summary of everything the Correctless workflow caught during this feature. This is the "look what I saved you" report.
Creates isolated Git worktrees for feature branches with prioritized directory selection, gitignore safety checks, auto project setup for Node/Python/Rust/Go, and baseline verification.
Executes implementation plans in current session by dispatching fresh subagents per independent task, with two-stage reviews: spec compliance then code quality.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
Generate a one-page summary of everything the Correctless workflow caught during this feature. This is the "look what I saved you" report.
/cdocs completes (end of a feature) — full summaryRead these files to build the summary. Skip any that don't exist.
.correctless/specs/{task-slug}.md) — check for rules added during review (rules not in the original draft).correctless/artifacts/qa-findings-{task-slug}.json) — issues caught during TDD QA.correctless/verification/{task-slug}-verification.md) — issues caught during verification.correctless/artifacts/tdd-test-edits.log) — tests modified during implementation.correctless/artifacts/audit-trail-{branch-slug}.jsonl) — every file modification with workflow phase and timestamp. Shows exactly which files were touched in which phases, without manual instrumentation.Read the workflow state file to get the task name, spec path, and branch. If no active workflow, ask the human which feature to summarize and look for the spec and artifacts by slug.
Read the spec file. Look for rules that were added during the review phase — these are rules the spec author didn't think of. Indicators:
guards_against: AP-xxx)If a research brief exists (.correctless/artifacts/research/{task-slug}-research.md), note what the research agent found (stale APIs, CVEs, deprecated patterns).
The test audit runs between RED and GREEN. Its findings are verbal (returned to the orchestrator) and may not be persisted as a file. Check:
Read .correctless/artifacts/qa-findings-{task-slug}.json. For each finding:
Read .correctless/verification/{task-slug}-verification.md. Extract:
For each issue caught, assess: would this have shipped to production without the workflow?
Count the "would have shipped" items. This is the headline number.
Print the summary to the conversation AND write it to .correctless/artifacts/summary-{task-slug}.md:
# Workflow Summary: {Feature Name}
**Branch:** {branch name}
**Duration:** {time from first to last commit}
**Spec:** {.correctless/specs/slug.md}
## What the Workflow Caught
### Review found ({N} issues): <!-- /creview (at standard intensity) or /creview-spec (at high+ intensity) -->
- {description of each issue added during review}
- {security checklist findings}
### Test Audit found ({N} issues):
- {test quality issues caught before implementation started}
### /ctdd QA found ({N} issues, {M} rounds):
- {each QA finding with instance + class fix}
### /cverify found ({N} issues):
- {verification findings}
## Research Insights (if research agent ran):
- {current best practices found, stale patterns avoided, CVEs dodged}
## Spec Updates During TDD:
- {if spec was revised mid-implementation, what changed and why}
## Stats
- **Total issues caught:** {N}
- **Would have shipped without workflow:** {M} ({percentage}%)
- **QA rounds:** {count}
- **Spec updates:** {count}
## What This Means
{1-2 sentences: "Without this workflow, {M} issues including {most critical} would have shipped to production."}
Use TaskCreate/TaskUpdate to show progress:
After generating: "Export this summary to include in your PR description: /export .correctless/artifacts/summary-{task-slug}.md"
/csummary needs QA findings from /ctdd).templates/redaction-rules.md first.