"Develop inside the OpenAI Agents Python repository safely,
Scanned 9/8/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill repo-development --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Repo Development?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-repo-development-492e7adc)More formats (shields.io, HTML) on the badges page.
---
name: repo-development
description: "Develop inside the OpenAI Agents Python repository safely,
including repo policies, compatibility reviews, tests, docs builds, snapshots,
and PR handoff expectations."
disable-model-invocation: true
metadata:
disco-role: operating
license: MIT
---
# Repo Development
Use this sub-skill when a task asks to modify this repository, add or update tests, touch `RunState` serialization, preserve public API compatibility, build docs, update snapshots, adjust `Makefile` / `pyproject.toml`, or prepare a PR-ready handoff.
## Start Here
- For maintainer workflow, code layout, docs scripts, and PR expectations, read [references/maintainer-workflows.md](references/maintainer-workflows.md).
- For test selection, quality commands, snapshots, docs builds, and compatibility checks, read [references/testing-and-quality.md](references/testing-and-quality.md).
- For common local failures and recovery paths, read [references/troubleshooting.md](references/troubleshooting.md).
- To map changed files to likely focused validation commands without running them, use [scripts/select_test_targets.py](scripts/select_test_targets.py).
## Routing Boundaries
- Stay here for repository policy, implementation planning, runtime-code edits, tests, snapshots, docs build workflow, generated reference docs, `RunState` schema policy, public constructor positional compatibility, and PR summary handoff rules.
- Route ordinary SDK runtime usage, `Agent`, `Runner`, `RunConfig`, streaming, and pause/resume behavior to ../core-runtime/SKILL.md.
- Route sandbox API design and safety boundaries to ../sandbox-agents/SKILL.md.
- Route provider/model transport and OpenAI-compatible provider code to ../models-providers/SKILL.md.
## Non-Negotiable Repo Rules
- Before changing runtime code, exported APIs, external config, persisted schemas, wire protocols, or user-facing behavior, perform an implementation-strategy pass and call out released compatibility risk early.
- Before marking runtime, test, example, build/test-config, or behavior-impacting docs changes complete, run the repo's code-change verification policy; the mandatory local order is `make format`, `make lint`, `make typecheck`, then `make tests`.
- Before the final handoff for eligible changes, prepare the PR draft summary block required by repository policy; this does not imply creating a branch, commit, push, or PR.
- Preserve public API positional compatibility: append optional dataclass fields/constructor parameters instead of inserting them, or add an explicit compatibility layer and regression tests.
- If serialized `RunState` shape changes, update `CURRENT_SCHEMA_VERSION`, keep `SCHEMA_VERSION_SUMMARIES` chronological and non-empty, and add round-trip/backward-read tests.
- Do not edit translated docs under generated language folders unless the task is explicitly about translation maintenance; source English docs and generated reference stubs are the normal edit surfaces.
## Quick Command Selection
From the repository root, pass changed paths to the helper:
```bash
python skills/openai-agents-python/sub-skills/repo-development/scripts/select_test_targets.py src/agents/run_state.py tests/test_run_state.py
```
Add `--json` for machine-readable suggestions. The helper only prints suggestions; it never runs tests, builds docs, mutates snapshots, or invokes networked services.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!