Skip to content

rampstackco/claude-skills / skills/oss-contribution-research

oss-contribution-research

Verified

Research and prepare an outside contribution to an open-source repo, then stop at a queue item for a human to approve: policy check, intake, duplicate check, reproduction, patch, the repo's own checks, an independent critic rerun, and a disclosed, hashed draft of the exact text to post. Use this skill whenever the user wants to contribute to someone else's repo, fix an upstream bug, pick up a good-first-issue, draft an AI-assisted pull request for an outside project, follow up on their own outside PR, or run a gated contribution pipeline. Triggers on open source contribution, contribute to this repo, upstream fix, outside PR, external PR, good first issue, fix this issue in their repo, contribution research, AI-assisted PR, disclose AI in PR, reply to review on my PR. Also triggers when the user names an issue in a repo they do not maintain and asks for a fix, even if they do not say contribute.

references/pr-description-template.md

View on GitHub
# Exact-text templates

Drafts for the "Exact text to post" block. This is the text that appears under the user's name, so it is short, specific, and disclosed in the target repo's own format. Take the disclosure template from the registry entry for the repo and the action type:

| Action | Registry key |
|---|---|
| Pull request | `pr_description` |
| Intake comment | `issue_comment` |
| Reply, including the 14-day ping | `review_reply` |
| Closing comment on the user's own pull request | `review_reply` |

Each template is `unknown` (stop), `none` (no disclosure line; the item's Disclosure text is `none`), or one line with `{slots}`. Fill every slot with what actually happened, put that exact line on its own line in the text, and copy it into the item's Disclosure field. A slot left as `<...>`, `{...}`, `TODO`, `TBD`, `e.g.` or `...` is refused.

For a repo whose policy class is `silent`, and for a repo whose policy makes disclosure optional, the registry usually carries these defaults:

- **Pull request description:** one prose line in the body, with no tool or vendor name and no footer:

  ```text
  Drafted with AI assistance and approved by me. Verification: {verification}.
  ```

  Write the verification in the passive voice and list only what was run, for example "the issue was reproduced, a regression test was added, and the full test suite passes locally." Do not say the user reproduced, tested, or reviewed the change personally.
- **Issue comment, reply and closing comment:** no disclosure line (template `none`). The user approves the text word for word and posts it under their own name. If a maintainer asks about AI use, the drafted reply answers directly and plainly.

**No AI attribution.** Some repos forbid attributing commits or pull request bodies to an AI tool. There, disclosure stays the prose line above, and no tool or model name, trailer, co-author line, or generated-by line appears in the body or in any commit message.

If the template is `unknown`, the repo has an explicit policy that does not cover this action type, or it has not been researched. Stop.

## Pull request (action: pr)

The first line is the title, then one blank line, then the body. Follow the repo's title convention if it has one, for example `fix(parser): ...`.

```text
[type(scope): what the fix does]

Fixes #[issue].

[What was wrong, in one or two sentences, from the user's side of the bug.]

[What the change does and why this approach, in two or three sentences.]

How to verify: [the failing test that now passes, or the reproduction steps].

[Disclosure in the registry's pr_description format.]
```

Rules:

- If the repo's pull request template has checkboxes or declarations, fill only those the user can truthfully make.
- Never tick a box that attests to a human review on the user's behalf. Leave it for the user to fill in through the approval step.
- No claims the diff does not support, such as "improves performance" without a measurement.
- No unrelated cleanup mentioned or included.

## Intake comment (action: intake)

```text
[One sentence: what was reproduced, and on which version or commit.]
[One sentence: the fix being considered, in plain terms.]
Would a pull request for this be welcome? [If the repo assigns issues: Could this be assigned to me?]

[Disclosure in the registry's issue_comment format.]
```

Keep it under 80 words. Maintainers read many of these. A reproduction makes the comment worth reading.

## Reply (action: reply)

```text
[Address each review point in order: what changed, or why not, in one or two sentences each.]
[Link the new commit.]

[Disclosure in the registry's review_reply format.]
```

Rules:

- Reply to the review. Do not reopen a design question the maintainer already settled.
- A reply that answers review with new commits starts a new review round. Name the new commit in the text.
- The poster only posts new comments; it never edits one. Where a repo requires agents to update an existing comment instead of posting another, the registry keeps that repo paused, and this skill stops. A lower reply frequency does not satisfy that rule.

## Ping (action: reply, purpose: ping)

Only for the user's own pull request, once, at 14 days without a maintainer response.

```text
[One sentence: a polite check-in on whether the change is still wanted.]
[Optional: one sentence offering to adjust scope, rebase, or split.]

[Disclosure in the registry's review_reply format.]
```

## Close (action: close)

Only for the user's own pull request, at 28 days with no response after the ping. The poster posts this comment, then closes the pull request.

```text
[One sentence: closing for now to keep the queue tidy.]
[One sentence: happy to reopen or rework if it becomes useful.]

[Disclosure in the registry's review_reply format.]
```

## Disclosure placement

| Pattern in the registry | Where it goes |
|---|---|
| Disclosure with extent | The final paragraph of the body, naming what the AI did (code, tests, description). |
| Fixed description line | Verbatim on its own line, with the agent's own name and link. |
| Environment disclosure | A short list: model, harness, harness version, plugins. |
| No AI attribution | The prose line, with no tool or model name. No trailer, no co-author line, no generated-by line, no badge, in the body or the commits. |
| Silent or optional policy | Pull request description: the prose line. Comments, replies and closing comments: none. |
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall