closed-loop-delivery
/closed-loop-delivery
Use when a coding task must be completed against explicit acceptance criteria with minimal user re-intervention across implementation, review feedback, deployment, and runtime verification.
- Author
- pinkpixel-dev ↗
- Repository
- skills-collection-1 ↗
- Licence
- Apache-2.0
- Last repository change
- 09.08.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 09.08.2026, without references to retired models or APIs.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
What it does and when to use it
Use when a coding task must be completed against explicit acceptance criteria with minimal user re-intervention across implementation, review feedback, deployment, and runtime verification.
Original author description
Use when a coding task must be completed against explicit acceptance criteria with minimal user re-intervention across implementation, review feedback, deployment, and runtime verification.
How it works
- Overview
- When to Use
- Required Inputs
- Issue Gate Dependency
- Default Workflow
- PR Comment Polling Policy
- Human Gate Rules (Must Ask)
- Iteration/Stop Conditions
- Output Contract
SKILL.md preview
--- name: closed-loop-delivery description: Use when a coding task must be completed against explicit acceptance criteria with minimal user re-intervention across implementation, review feedback, deployment, and runtime verification. risk: safe source: community date_added: "2026-03-12" --- # Closed-Loop Delivery ## Overview Treat each task as incomplete until acceptance criteria are verified in evidence, not until code is merely changed. Core rule: **deliver against DoD (Definition of Done), not against code diff size.** ## When to Use Use this skill when: - user gives a coding/fix task and expects end-to-end completion - task spans code + tests + PR comments + dev deploy + runtime checks - repeated manual prompts like "now test", "now deploy", "now re-check PR" should be avoided Do not use this skill for: - pure Q&A/explanations - prod deploy requests without explicit human approval - tasks blocked by missing secrets/account access that cannot be inferred ## Required Inputs Before execution, define these once: - task goal - acceptance criteria (DoD) - target environment (`dev` by default) - max iteration rounds (default `2`) If acceptance criteria are missing, request them once. If user does not provide, propose a concrete default and proceed. ## Issue Gate Dependency Before execution, prefer using `create-issue-gate`. - If issue status is `ready` and execution gate is `allowed`, continue. - If issue status is `draft`, do not execute implementation/deploy/review loops. - Require user-provided, testable acceptance criteria before starting execution. ## Default Workflow 1. **Define DoD** - Convert request into testable criteria. - Example: checkout task DoD = "checkout endpoint returns a valid, openable third-party payment URL in dev". 2. **Implement minimal change** - Keep scope tight to task goal. 3. **Verify locally** - Run focused tests first, then broader checks if needed. 4. **Review loop** - Fetch PR comments/reviews. - Classify valid vs non-actionable. - Fix valid items, re-run verification. 5. **Dev deploy + runtime verification** - Deploy to `dev` when runtime behavior matters. - Verify via real API/Lambda/log evidence against DoD. 6. **Completion decision** - Only report "done" when all DoD checks pass. - Otherwise continue loop until pass or stop condition. ## PR Comment Polling Policy Avoid noisy short polling by default. Use batched windows: - **Round 1:** wait `3m`, collect delta comments/reviews - **Round 2:** wait `6m`, collect delta again - **Final round:** wait `10m`, collect all remaining visible comments/reviews At each round: - process all new comments in one batch - avoid immediate re-poll after each single comment - after the `10m` round, stop waiting and proceed with all comments visible at that point If CI is still running, align polling to check completion boundaries instead of fixed rapid polling. ## Human Gate Rules (Must Ask) Require explicit user confirmation for: - production/staging deploy beyond agreed scope - destructive operations (history rewrite, force push, data-destructive ops) - actions with billing/security posture changes - secret values not available in repo/runtime - ambiguous DoD that materially changes outcome ## Iteration/Stop Conditions Stop and escalate with a concise blocker report when: - DoD still fails after max rounds (`2` default) - external dependency blocks progress (provider outage, missing creds, account permission) - conflicting review instructions cannot both be satisfied Escalation report must include: - what passed - what failed - evidence (commands/logs/API result) - smallest decision needed from user ## Output Contract When claiming completion, always include: - acceptance criteria checklist with pass/fail - commands/tests run - runtime evidence (endpoint/Lambda/log key lines) - PR status (new actionable comments count) Do not claim success without evidence.
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · pinkpixel-dev/skills-collection-1 | 100 | Apache-2.0 |
| FrancoStino/opencode-skills-collection | 100 | MIT |
| LiHongwei-cn/lihongwei-cn | 100 | MIT |
Identical content appears in 1 other GitHub folders. We show one copy with the highest score.
Database snapshot: 13.11.2026 · package version 2981317768b6 (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
- 90acceptance-orchestrator
Use when a coding task should be driven end-to-end from issue intake through implementation, review, deployment, and acceptance verification with minimal human re-intervention. - 100acceptance-orchestrator
Use when a coding task should be driven end-to-end from issue intake through implementation, review, deployment, and acceptance verification with minimal human re-intervention. - 90acceptance-orchestrator
Use when a coding task should be driven end-to-end from issue intake through implementation, review, deployment, and acceptance verification with minimal human re-intervention. - 90delivery
Carry a software task end to end, from an isolated worktree through implementation, verification and delivery to a resumable handoff. Use for substantial feature, bug-fix or refactor work, especially when the user hands off a task and expects a finished, verified result without supervising it. - 100verify-acceptance-criteria
Use when asked to "verify acceptance criteria", "check criteria", or "verify the implementation against the plan" during active feature development. Do NOT trigger for issue management (close issue, comment), worktree cleanup, or post-merge/post-PR operations. - 100subagent-driven-development
Use when executing an implementation plan in the current session (the task-plan implement engine; the acceptance-criteria ledger engine is /skill:loop-implementation)