open-pr
Read before any PR create or edit: description, screenshots, and evidence.
Skill by gutierrezje
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/gutierrezje-open-pr
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: writing
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/open-pr
What's inside
- Skill name
open-pr- Description
- Read before any PR create or edit: description, screenshots, and evidence. Repairs: babysit-pr.
- License
- MIT
- Instructions
- 251 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- openai.yaml · agents/
SKILL.md (first lines)
# Open PR
Create a PR that a reviewer can understand and start reviewing without
reconstructing the work from the branch. Finish with the PR open for review by
default.
Review-ready means the PR's scope and evidence support meaningful review, with
remaining limitations explicit and no review blockers. It does not mean every
hosted check has settled or the branch has earned a current-head acceptance
verdict. `$babysit-pr` owns full-branch review, repeated repair coverage,
review feedback, hosted-check
convergence, and the final `READY`, `NOT READY`, or `INCONCLUSIVE` verdict.
Read and follow the [GitHub transport contract](../jesus-mode/references/codex-compatibility.md#github-transport) for every PR operation.
## Authorization
An explicit invocation of `$open-pr`, or a direct request to make, create, open,
update, or prepare a PR, authorizes commits, a non-force push to the intended
branch, creating or updating that PR, and marking it ready for review, including
a draft from an earlier workflow. Honor an explicit request to keep the PR in
draft or limit the task to a narrower edit. Existing draft state alone does not
require another user prompt to open it for review. Automatic model selection
does not grant publishing authority. Keep force pushes, remote history rewrites,
retargeting, public comments, review-thread actions, closing, and merging behind
separate user authorization.
## Review blockers
Keep the PR in draft for a known required-check failure, an unmet explicit
verification requirement, or an evidence gap that prevents assessing a material
behavior or usability claim. Name the requirement or claim each blocker affects.
A missing screenshot blocks only when explicitly required or needed to
substantiate a material claim. If an optional capture or upload fails, omit the
image and disclose the limitation when the remaining evidence suffices.
Additional captures and presentation polish do not block meaningful review.
Repository and linked-issue acceptance requirements remain binding, including
specified native journeys. Record any user-approved alternative evidence or
scope change in the PR. Reporting a required journey as unverified does not
waive it, and this skill's evidence flexibility does not authorize that change.
## Workflow
1. Resolve the repository, base branch, head branch, linked issue, worktree
state, and any existing PR. Preserve unrelated work.
2. Inspect the complete intended diff against the base once as a bounded PR
preflight. Include staged and unstaged work plus enough surrounding code to
confirm the scope, describe it accurately, remove accidental churn, and fix
clear in-scope problems. This is preparation, not an exhaustive review loop.
3. Run the checks required by the change, repository, and linked issue. Verify
the behavior on the most faithful practical surface.
4. Gather direct evidence for every material claim in the PR. Use the evidence
rules below. Reuse evidence collected during implementation after confirming
it still covers the prepared change. Collect missing or stale evidence needed
to close a review blocker. If a before image is needed, reproduce the base in
an isolated checkout when practical or use the labeled historical baseline
described below. Narrow or remove unsupported claims and disclose remaining
gaps; apply the review-blocker criteria before deciding draft state.
5. Commit coherent implementation units using repository conventions. Apply
the evidence storage rules below before staging artifacts. Reuse current
review results when they already exist, but do not
start or extend a model-review coverage chain solely to open the PR.
6. Push non-destructively to the intended head branch. Confirm the remote OID
matches the prepared local OID.
7. Create or update the PR. Write the body from the prepared diff and collected
evidence, including the [model contributions](#model-contributions) record,
not from memory or the commit titles alone. Create new PRs ready
for review unless the user requested a draft or a review blocker remains.
8. Read the remote PR back through the CLI or GitHub plugin. Confirm its head SHA
matches the prepared local head and its title, body, and links are correct.
Verify native screenshot attachments and inline image rendering through the
GitHub transport contract. Fix leftover local paths, repository-image targets,
and link-only substitutes for required screenshots before calling the PR
review-ready.
9. For any remaining draft, run `gh pr ready` or the GitHub plugin equivalent
after every review-blocking evidence gap is closed, unless the user requested
draft state or a narrower edit. Pending hosted checks do not block this
transition, but a known required-check failure on the current head does. Read
the PR back and confirm the same head is now open for review.
10. Return the PR URL, head SHA, verification and artifact summary, current check
state, and any evidence gaps. Begin `$babysit-pr` only when the user
explicitly asked to babysit, shepherd, or finish the PR. If Babysit is
already active, return to that run.
Keep this workflow bounded. Use focused verification after any preflight fix,
then finish the PR presentation. Repeated model review, repair-delta review,
review-thread triage, and waiting for checks to converge belong to
`$babysit-pr`.
## PR body contract
Use these sections:
```markdown
## What
<The behavior or result that changed.>
## How
<The implementation approach and important design choices.>
## Why
<The problem, constraint, or decision that motivated the change.>
## Evidence
<Direct proof for the claims above.>
## Model contributions
| Role | Model | Scope and result |
| --- | --- | --- |
| Implementation | <model> | <Changes authored, including repairs.> |
| Review | <model> | <Review type, covered revision or range, and result.> |
| QA | <model> | <User journeys exercised, revision or build, and result.> |
| Verification | <model> | <Tests or checks run, covered revision, and result.> |
| PR preparation | <model> | <PR creation or description update.> |
```
Add risks, rollout notes, or reviewer guidance when they help. Keep the fiveFrom gutierrezje/jstack, MIT.
More from gutierrezje/jstack
- 0.567score
create-verification-skill
gutierrezje
Create project-local behavior verification skills and adapters when asked to build repeatable UI, CLI, mobile, or service checks.
SkillAFreshness: AClaude Code +9 - 0.564score
architect
gutierrezje
Use when asked to design module boundaries, types, or interfaces before implementation.
SkillAFreshness: AClaude Code +9 - 0.564score
babysit-pr
gutierrezje
Babysit a PR to ready: review, repairs, CI, and a current-head verdict.
SkillAFreshness: AClaude Code +9 - 0.564score
jesus-mode
gutierrezje
Default workflow for features, bug fixes, investigations, issue triage, and PRs.
SkillAFreshness: AClaude Code +9 - 0.561score
maintain-verification-skill
gutierrezje
Audit and repair a project's verification skills, executable adapters, feature coverage, and receipts against current source and live behavior.
SkillAFreshness: AClaude Code +9 - 0.558score
how
gutierrezje
How code works: architecture, runtime flow, ownership, or where a change belongs.
SkillAFreshness: AClaude Code +9 - 0.558score
arena
gutierrezje
Use when asked to compare parallel candidate implementations or designs and combine the strongest results.
SkillAFreshness: AClaude Code +9 - 0.556score
pstack-teach
gutierrezje
Explain a body of work plainly so a person actually understands it.
SkillAFreshness: AClaude Code +9 - 0.553score
setup-jstack
gutierrezje
Configure Jstack model, effort, and fan-out routes.
SkillAFreshness: AClaude Code +9 - 0.553score
automate-me
gutierrezje
Create or revise a personal mode skill when asked to capture working preferences as agent instructions.
SkillAFreshness: AClaude Code +9 - 0.553score
recall
gutierrezje
Recover recent working context when asked to resume prior work or recall earlier tasks and decisions.
SkillAFreshness: AClaude Code +9 - 0.553score
figure-it-out
gutierrezje
Use for a large migration or multi-part task that needs a custom execution playbook.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 20, 2026 (9 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/gutierrezje/jstack
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from gutierrezje
- 0.567score
create-verification-skill
gutierrezje
Create project-local behavior verification skills and adapters when asked to build repeatable UI, CLI, mobile, or service checks.
SkillAFreshness: AClaude Code +9 - 0.564score
babysit-pr
gutierrezje
Babysit a PR to ready: review, repairs, CI, and a current-head verdict.
SkillAFreshness: AClaude Code +9 - 0.564score
architect
gutierrezje
Use when asked to design module boundaries, types, or interfaces before implementation.
SkillAFreshness: AClaude Code +9 - 0.564score
jesus-mode
gutierrezje
Default workflow for features, bug fixes, investigations, issue triage, and PRs.
SkillAFreshness: AClaude Code +9 - 0.561score
maintain-verification-skill
gutierrezje
Audit and repair a project's verification skills, executable adapters, feature coverage, and receipts against current source and live behavior.
SkillAFreshness: AClaude Code +9 - 0.558score
arena
gutierrezje
Use when asked to compare parallel candidate implementations or designs and combine the strongest results.
SkillAFreshness: AClaude Code +9
Cite this page
open-pr skill. Agent Mart. https://agentmart.dev/skills/gutierrezje-open-pr (last updated 2026-10-24)