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/repro-record-template.md

View on GitHub
# Repro record

The evidence a `pr` item carries for gate 8. It records what actually ran, with numbers a second person can check without trusting the first. An intake item does not need one: it asks before any patch, and its Summary says what was confirmed from the code. A `pr` item without a repro record is not a `pr` item.

## The browser rule

A bug that shows in a browser is reproduced in a browser. That needs a connected browser tool in the session: one that loads the page, runs the assertion and captures the screen. Reading the renderer's source and the fixture files can explain a bug; it does not reproduce it. Without a connected browser, the run may still file intake items, and the `pr` item waits for a session that has one.

The same rule applies to any surface the agent cannot drive from the session: a mobile build, a desktop app, a device. If the assertion cannot run, the record stays empty and the item stays intake.

## Fields

Every field is filled. A field that does not apply says why, in words, not with a dash.

```
target:       owner/repo#n, in a code span when the record is pasted into a queue item
surface:      where the assertion ran: a local build at http://localhost:<port>, a CLI command, a test runner, a hosted page
steps:        numbered; one navigation or one in-page action per line; the exact file, URL or input each step used
assertion:    one measurable check (a selector count, an exit code, a returned value, a log line), with
              expected: the value a fixed build gives
              observed: the value the unpatched build gave
control:      the one change that flips the observed value to the expected one (a patch, a config line, an input edit), and the value after it
second case:  a second input that shows the same failure, with its expected and observed values
regression:   an input that must not change: its value before and after the control
environment:  browser and version, runtime and version, the repo's commit SHA, the OS
evidence:     paths to the screenshot or GIF, and the raw output of the assertion before and after
```

## Rules

- **Observed means run.** Every number in the record came from the assertion running on the surface named. A value worked out from the source is a prediction and goes in the Summary, labeled as such.
- **The control is one change.** If flipping the result took two changes, the record names both and the patch is probably too big for the size gate.
- **Second case and regression are different inputs.** The second case fails the same way; the regression input does not change. One input cannot be both.
- **Environment names a commit.** The repo's SHA at which the record was taken, so the critic can rerun it at the same head.
- **Evidence is a file, not a sentence.** The critic reruns the assertion and compares its output with the files named here.
- **Paste the record, not a summary of it.** The `pr` item's repro gate line names the record's path and the before and after values; the record itself goes into the item in full, as a fenced block at the end of the Summary section after the Files touched line. The poster reads a fenced block as content, so the record never breaks its parse.

## Filled example

A fictional repo, `example-org/chartkit`, ships a hosted viewer for saved chart files. Issue `example-org/chartkit#88` reports that files written before schema 3.2 show no rows in the Timing panel, although the data is in the file.

```
target:       `example-org/chartkit#88`
surface:      local build of the viewer at http://localhost:8080, served from the repo's `viewer/dist`
steps:        1. Open http://localhost:8080/?file=http://localhost:8080/fixtures/chart-3.0.json
              2. Wait for the Timing panel heading to render
              3. In the console, run document.querySelectorAll('#timing .ck-row').length
assertion:    document.querySelectorAll('#timing .ck-row').length
              expected: 12 (the 12 timing entries in chart-3.0.json)
              observed: 0
control:      in viewer/src/compat.js, add groups.timing = {title: 'Timing'} when the file has timing entries and no timing group; after it, the count is 12
second case:  fixtures/chart-2.8.json, expected 9, observed 0; after the control, 9
regression:   fixtures/chart-3.4.json (current schema), 15 before and 15 after the control
environment:  Chrome 131.0.6778.86, Node 22.11.0, chartkit at 4f2c9e1b7d3a0c5e8b6f1a2d9c4e7f0b3a6d8e1c, Windows 11
evidence:     repro/chartkit-88/timing-before.png, repro/chartkit-88/timing-after.gif,
              repro/chartkit-88/assertion-before.json ({"chart-3.0": 0, "chart-2.8": 0, "chart-3.4": 15}),
              repro/chartkit-88/assertion-after.json ({"chart-3.0": 12, "chart-2.8": 9, "chart-3.4": 15})
```

In the `pr` item, the repro gate line reads `- repro: PASS | repro/chartkit-88/record.md, assertion 0 before and 12 after at 4f2c9e1` and the record above sits in a fenced block at the end of the Summary section.
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