REPO PACKAGE
openai openai-agents-python
OFFICIAL13skills
3ready to use
LABSCO SUMMARY
Our capture has fourteen skill entries here but only thirteen carry a description; the fourteenth ships with no frontmatter text to summarize, so it is left out of the breakdown below. Of the thirteen documented ones, eight are process skills for people shipping this exact package: code-change-verification and pr-draft-summary run the required lint, type, and test checks before a PR is proposed; maintainer-review and implementation-strategy are the rubric a maintainer applies to an incoming issue or feature request; docs-sync, final-release-review, runtime-behavior-probe, and test-coverage-improver round out the release and QA loop.
The remaining five are not maintenance work, and only one of them is close to the SDK's own subject matter: openai-knowledge retrieves current OpenAI API documentation, and playwright is a generic terminal browser-automation skill with no tie to this package. csv-workbench summarizes CSV files, credit-note-fixer patches one specific formatting bug and reruns one specific test, and prior-auth-packet-builder assembles health-insurance prior-authorization packets — none of which have anything to do with the OpenAI Agents SDK. Someone wanting to learn the SDK's own agent, tool, and handoff model should read its documentation and examples directory, not these skills.
READ THE FULL ANALYSIS
What it costs you to start. Nine of the fourteen need a local tool already on the machine — the test runner, the linter, git — one wants a stored account key, one talks to an MCP server, and three are ready with no setup. None of that is unusual for a maintainer checklist; it assumes you already have this repository cloned with its own dev environment working.
ALSO IN THIS PACKAGE
PROMPT TEMPLATES
4 itemsPROMPT TEMPLATE
memory_consolidation_prompt
PROMPT TEMPLATE
memory_read_prompt
PROMPT TEMPLATE
rollout_extraction_prompt
PROMPT TEMPLATE
rollout_extraction_user_message
WHAT'S INSIDE
13 showing · 13 totalReady to use
Nothing else to set up — install it and go.
13 of 13
SETUP
Showing all 13 skills
code-change-verification
Nothing here is called finished until the code has been tidied, checked over for mistakes, and put through the whole test suite — in that order, stopping at the first thing that fails.
credit-note-fixer
You find and fix a formatting bug in how this project generates credit notes, then confirm the fix with the project's own test script.
csv-workbench
Hand it a table of data and it comes back with the numbers you asked for — counts, sums, averages — saying plainly where it had to guess about gaps or bad rows.
docs-sync
The documentation drifts behind the code; this finds where — features nobody wrote up, defaults that quietly changed — and, with your go-ahead, edits the English pages and rebuilds the site.
final-release-review
Should this go out, and as a small update or a feature release? It reads everything that changed since the last version, ranks what could bite, and says stop or go.
implementation-strategy
Before writing the fix, write down what must work and what you will deliberately refuse — so the change stays small instead of growing with every round of review comments.
maintainer-review
Decides what to do with an incoming bug report or contribution by asking first whether anyone actually needs it — the fix the reporter proposed is a suggestion, not the requirement.
openai-knowledge
Fetches the page from OpenAI's own documentation and answers from what it actually says, rather than from what it half-remembers about parameters and limits.
playwright
Takes a picture of a web page sitting in a folder on your computer — opening it in a real browser with no window on screen and no web server involved.
pr-draft-summary
When the work is done it drafts the title, description and branch name for you to paste into GitHub, whether or not you remembered to ask for one.
prior-auth-packet-builder
Reads a patient's case notes against the insurer's own policy documents and writes up what they say about approval and referrals, plus a checklist for the person who signs off.
runtime-behavior-probe
Settles an argument about how the software really behaves by running it, not by reading it — and nothing runs until you have seen the exact command and approved it.
test-coverage-improver
Which parts of the code has nobody tested? This answers that from real measurement rather than a target percentage, and — once you agree — writes the missing tests.
No skill in this package matches that.
HOW TO GET IT
WHOLE PACKAGEInstall the whole package (13 skills)
npx skills add openai/openai-agents-pythonSINGLE SKILLOr install a single skill
npx skills add openai/openai-agents-python --skill <name> --full-depthPick the skill name from the Skills tab — each entry there installs independently.