finish-task
/finish-task
Use when implementation is done and the task is ready to commit and PR. Runs simplify → review → score → commit → PR in sequence. Triggers on "finish task", "ready to PR", "wrap up", "close out", "done implementing", "submit PR".
- Author
- digithings-ai ↗
- Repository
- digithings ↗
- Licence
- MIT
- Last repository change
- 12.11.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 12.11.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Use when implementation is done and the task is ready to commit and PR. Runs simplify → review → score → commit → PR in sequence. Triggers on "finish task", "ready to PR", "wrap up", "close out", "done implementing", "submit PR".
Triggers from the description
Original author description
Use when implementation is done and the task is ready to commit and PR. Runs simplify → review → score → commit → PR in sequence. Triggers on "finish task", "ready to PR", "wrap up", "close out", "done implementing", "submit PR".
How it works
- Sequence
- Failure paths
SKILL.md preview
--- name: finish-task description: Use when implementation is done and the task is ready to commit and PR. Runs simplify → review → score → commit → PR in sequence. Triggers on "finish task", "ready to PR", "wrap up", "close out", "done implementing", "submit PR". --- # Finish a task Run this close-out sequence **after** implementation is complete. All steps must pass before the PR is created. ## Sequence ### 1. Simplify Review the changed files for: - **Reuse** — any new function that duplicates existing utilities? Flag it. - **Quality** — redundant state, copy-paste blocks, stringly-typed code, unnecessary nesting (3+ levels), comments that explain WHAT instead of WHY. - **Efficiency** — redundant computations, missed concurrency (independent ops run sequentially), overly broad file reads. Fix every non-trivial finding before continuing. ### 2. Review Review against each rubric in `docs/scoring/`, using whatever review tooling your harness already provides (Cursor Bugbot, an in-session `/review`-style fan-out, or a PR-review plugin) rather than a bespoke subagent: | Dimension | Min | Rubric | |---|---|---| | Security | ≥ threshold | `docs/scoring/SECURITY.md` | | Quality | ≥ threshold | `docs/scoring/QUALITY.md` | | Optimization | ≥ threshold | `docs/scoring/OPTIMIZATION.md` | | Accuracy | ≥ threshold | `docs/scoring/ACCURACY.md` | Thresholds are in `agents.yml` → `scoring_thresholds`. Address all findings before continuing. ### 3. Score ```bash make score ``` Must exit 0. If any dimension fails: read the rubric, apply the narrowest fix, re-stage, re-run. If score fails twice, stop and escalate — do not open the PR with known violations. ### 4. Commit ```bash make commit MSG="type(component): short description (#N)" ``` Conventional commit format required. Include the issue number. ### 5. PR ```bash make pr ``` Fill in the PR template's self-score checklist honestly. Check any applicable human gate boxes. Do not merge without human review if a gate is triggered. ## Failure paths - **Simplify finds issues** → fix, re-stage, continue to review. - **Review finds issues** → fix, re-stage, re-run score. - **Score fails once** → fix the narrowest failing criterion, re-stage, re-run. - **Score fails twice** → stop. Note the blocker. Escalate to a human. - **`make pr` fails** → check `gh auth status`. Push and PR manually if needed.
Versions
The database has no other skills named finish-task.
Database snapshot: 13.11.2026 · package version ac686a2ce1f0 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 90go
Ship what built. Run verify -> self-review -> simplify -> commit-push-pr -> monitor CI -> fix -> done. Use when implementation done, ready to launch. - 100finish
Gates 6 and 7 for a finished task - self-review, decisions, the retro, the metrics line, committing the record files by path, the merge with agent-steddy record finish, releasing what the task started, and the report. Use when a task passed gate 5 and is ready to record and deliver. - 90finish-work
Use when a change is ready ("finish", "wrap up", "we're done", "open the PR"). Checks the docs moved with the change, runs the full gate with the plugin's types, tries in a real session what no test can reach, commits, opens a draft PR, reviews it with review-code until it is… - 90finish-work
Use when a change is ready ("finish", "wrap up", "we're done", "open the PR"). Checks the docs moved with the change, runs the full gate with the plugin's types, tries in a real session what no test can reach, commits, opens a draft PR, reviews it with review-code until it is… - 100pr-ready
Take a Hostel pull request to "ready to merge". Use it every time you create a pull request, and when the owner asks to finish, review, or babysit one. It runs a local subagent review, fixes the findings, follows the reviewers' comments and CI, and tells the owner when the pull request is ready. - 86done
Close a coding task. Use it when the person says the task is done, asks to wrap up or close it, or before delivering a medium or large task.