Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add mblode/agent-skills --skill pr-creatorgit clone --depth 1 https://github.com/mblode/agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/mblode/agent-skills/pr-creator)<a href="https://agentmods.dev/skills/mblode/agent-skills/pr-creator"><img src="https://agentmods.dev/badge/skills/mblode/agent-skills/pr-creator/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/mblode/agent-skills/pr-creator"><img src="https://agentmods.dev/badge/skills/mblode/agent-skills/pr-creator.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
- Socket pass
- Snyk pass
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 36 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00055 | $0.02937 |
| Opus 5.5 | $0.00022 | $0.01175 |
| Sonnet 5.5 | $0.00011 | $0.00587 |
| Haiku 4.5 | $0.00006 | $0.00294 |
Grade A, and why
pr-creator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 160 lines — stays where its author put it; the contents beside it link to each section on GitHub.
pr-creator
Write PR descriptions like a developer posting in Slack, not an AI summarizing a diff. The reviewer has none of your context, so the body answers what the diff cannot: why, what it costs if wrong, and where you want their view. The diff answers what changed.
- IS: creating or updating a GitHub PR's title, body, draft state, and reviewers, attaching visuals for a UI change, plus the commit restructuring and review path that make a large diff readable.
- IS NOT: changing the code in the diff or reviewing it for bugs (use
tidy), watching CI and review comments after the PR exists (usepr-babysitter), or cutting npm releases (useautoship).
Reference Files
| File | Read When |
|---|---|
references/pr-polish.md |
Commits contain fixup, WIP, or "address review" noise, the diff exceeds 500 lines, or the user asks to polish, restructure, squash, or split the PR |
Workflow
Copy this checklist to track progress:
PR creation progress:
- [ ] Inspect the current branch and diff against the existing PR base, or the discovered default branch merge-base
- [ ] Existing PR? gh pr view --json url,state,isDraft (state OPEN means edit, not create)
- [ ] Linear ID from branch, commits, prompt, or issue link; no ID means no prefix
- [ ] Noisy commits or a >500-line diff? references/pr-polish.md, before pushing
- [ ] PR template present? Read it (paths under Templates)
- [ ] Draft title and body against Rules and Anti-patterns
- [ ] git push -u origin HEAD
- [ ] gh pr create (or gh pr edit); then gh pr view --json url,title and return the URL
- [ ] UI change with a video or screenshots? gh pr edit <n> --attach <file> after the final body, then gh pr view <n> --json body
Do not ask the user to approve the description first; the point is speed. They can ask for a rewrite and you run gh pr edit.
Rules
- Title. With a Linear ID:
ABC-123: Add auth flow. Without one:Add auth flow. Under 60 characters, no trailing period. If the repo lints PR titles (asemantic-pull-requestor commitlint workflow under.github/workflows/), use its shape and put the ID at the end:feat: add auth flow (ABC-123). Linear links on the ID wherever it sits in the title. - Body. One short paragraph: what changed and why it matters. Length follows the change; a one-line body is fine for a one-line fix.
- No fake why. If the reason is not in the prompt, the Linear issue, the branch, the commits, or the diff, leave it out rather than inventing one.
- Risk and input only when real. One
Risk:line for migrations, billing, auth, permissions, irreversible writes, wide blast radius, or a subtle behavior change. OneInput wanted:line when a decision in the diff is still open and the reviewer's view would change it; name the decision, never "thoughts welcome". Otherwise nothing. - Testing only if real. Say what you ran only if you ran it. Never a
Test plansection, never checkboxes. - Partial work. If this PR does not finish the Linear issue, keep the ID out of the title and write
Part of ABC-123in the body. Linear's default automation moves a linked issue to Done when the PR merges; the non-closing magic word links without closing. - Draft and reviewers only when asked.
--draftfor "draft" or "WIP";gh pr readypromotes it later,gh pr ready --undodemotes.--reviewer alice,org/teamonly for people the user named; CODEOWNERS already requests owners. - Stop after the useful content. The harness's
attributionsetting owns the PR footer (Generated with Claude Code, session link). Never hand-write one, and never strip one the harness or a repo policy appends. Everything above the footer is this skill's; its one-paragraph body replaces the harness's defaultSummaryandTest plansections.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +20 lines 23b05abaae25
- 1mo ago Changed · -1 lines · -21 tokens per session 296010427cf5
- 1mo ago Changed · +1 tokens per session 9ce5ec7f80d1
- 1mo ago Changed · +3 lines 639b657eb634
- 1mo ago Changed 43710f8fc48c
- 2mo ago Changed · -4 lines · -141 tokens per session 233343941999
- 2mo ago First seen · 142 lines · 216 tokens per session scan A 5d6f2f85676b
pr-creator is a skill published in the GitHub repository mblode/agent-skills (191 stars, last pushed today), licensed MIT. It adds 55 tokens to every session and 2,937 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
pr-creator compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| pr-creatormblode/agent-skills | Skill | 191 | today | MIT | A | 55 | Claude Code, Codex |
| githubNousResearch/hermes-agent | Skill | 262k | today | MIT | A | 19 | Claude Code, Codex |
| pr-takeoverclosedloop-ai/claude-plugins | Skill | 124 | today | Apache-2.0 | A | 64 | Claude Code |
| pr-descriptionpetekp/claude-code-setup | Skill | 49 | 14d ago | MIT | A | 50 | Claude Code, Codex |
| diffctxnikolay-e/diffctx | Skill | 6 | 3d ago | Apache-2.0 | A | 57 | Claude Code |
| github-code-reviewgraniet/kheish | Skill | 212 | 2d ago | Apache-2.0 | A | 41 | Claude Code, Codex |
| github-pr-workflowStarryCod/cogitum | Skill | 12 | 5mo ago | MIT | D | 20 | Claude Code, Codex |