accountable-engineerin…
Use explicit human checkpoints for architecture, security, rollout, or operational decisions.
Write a fast-to-review pull request body with a visual summary, before/after evidence, and merge-risk analysis.
$ npx -y skills add jellydn/my-ai-tools --skill pr --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/prContext preview
The summary Claude sees to decide when to auto-load this skill.
Write a fast-to-review pull request body with a visual summary, before/after evidence, and merge-risk analysis.
name: pr description: "Write a fast-to-review pull request body with a visual summary, before/after evidence, and merge-risk analysis." version: 1.0.0 author: my-ai-tools license: MIT compatibility: cline, claude, opencode, amp, codex, gemini, cursor, pi hint: Use when preparing or updating a pull request for human review. metadata: audience: all workflow: git related_skills: [draft-pull-request, code-review, commit-atomic]
Prepare a pull request body that lets a human understand the change quickly, verify that it works, and spend review time where risk is highest. This skill improves the review narrative; it does not replace code review or validation.
Use when:
Load `draft-pull-request` for repository preflight, branch handling, and `gh` publication rules. Reuse an existing PR instead of opening a duplicate.
Before writing, gather:
Never claim a test, screenshot, or manual check that was not actually run.
Use these sections unless the repository's template requires equivalent headings:
## Summary <the smallest visual that explains the change> ## Evidence - **Before:** <failing test, old output, old screenshot, or baseline behavior> - **After:** <passing test, new output, new screenshot, or verified behavior> ## Merge Danger **Door:** <two-way or one-way> **Blast Radius:** <one-word scope> <what can break, what rollback restores, and where careful review matters>
Choose one small visual, not a prose dump:
Example:
implement-spec
read spec + ticket graph
frontier -> isolated workers
ticket implementation + tests
merge verified commits
code-review integration branchKeep only the files, calls, states, and boundaries needed to understand the change. Use the domain terms from `GLOSSARY.md` when the repository has one.
Evidence should be execution-based whenever possible:
- **Before:** `pytest tests/test_export.py::test_csv` — failed: endpoint missing - **After:** `pytest tests/test_export.py::test_csv` — passed - **Full check:** `pytest -q` — 142 passed
For visual work, prefer screenshots. For bug fixes, show the original reproduction and the regression check. If there is no meaningful before state, say so and provide the strongest available verification instead of manufacturing one.
For skill changes, show the learning loop rather than only the edited Markdown:
- **Input:** <stable task or fixture> - **Before:** <output from the previous skill> - **Human edit:** <decision the user made and why> - **After:** <output after updating the skill> - **Rerun:** <same input compared against the intended result>
Prefer a decision rule with boundaries over a literal preference. If the lesson can be checked mechanically, include the lint/test/hook that now enforces it.
Classify the door:
Name the blast radius plainly: `single-command`, `one-service`, `shared-library`, `data`, `all-users`, or another accurate scope. Mention rollback steps, migration order, compatibility concerns, and the most important review seam.
After writing and checking the body:
1. Use the `draft-pull-request` workflow to create or update the PR with `--body-file`. 2. Verify the returned PR URL, title, base, head, and draft state. 3. Report the exact validation results and any environment-only blockers.
1. Writing a paragraph where a small diagram would make the change obvious. 2. Listing tests without showing the relevant before/after result. 3. Saying “low risk” without identifying reversibility or blast radius. 4. Treating a revert as complete rollback for a migration or external side effect. 5. Describing intended behavior instead of the behavior actually verified. 6. Replacing repository-specific PR headings or checked items without reading its template. 7. Opening a second PR for a branch that already has one.
Portable, source-controlled configuration for AI coding assistants. my-ai-tools installs one consistent set of agent guidelines, skills, commands, hooks, MCP servers, and tool settings across supported AI coding assistants.
Use explicit human checkpoints for architecture, security, rollout, or operational decisions.
Record architecture decisions with ADRs — captures why, alternatives considered, and consequences
Use when monitoring an open GitHub PR for CI failures, review feedback, mergeability, and safe retries or fixes.
Inspect history and architecture for hidden constraints before risky cross-cutting changes.
Build an interactive report or experiment when the user asks to explore model capabilities.
Audit a diff for structural maintainability and unnecessary complexity.