The Agent Skills Directory

[SAFE]: The skill documents a secure release process using PyPI Trusted Publishing (OIDC), which effectively eliminates the need for managing long-lived API tokens or hardcoded credentials.
[SAFE]: A dedicated 'Guardrails' section provides clear instructions to the agent to avoid high-risk operations, such as creating tags or publishing development builds to PyPI, unless explicitly directed by the user.
[SAFE]: Local preflight commands utilize standard, well-known development tools (uv, ruff, pytest) to validate internal scripts and documentation, with no evidence of malicious command injection or remote code execution.
[SAFE]: The use of internal scripts for versioning and build preparation is consistent with the stated development workflow of the Switchyard project and does not involve suspicious external network calls or data exfiltration.

publish-package