WorkflowPromptGuard v0.2.0

Highlights

• Scan public GitHub repositories from a structured issue form.
• Get a deterministic security report plus an optional, visibly labeled GitHub Models explanation.
• No separately stored AI API key: the workflow uses its short-lived GITHUB_TOKEN and free rate-limited GitHub Models access.
• Target repositories are read at an immutable commit and never cloned or executed.
• AI input is limited to rule-catalog aggregates; raw issue and workflow content is excluded.

See CHANGELOG.md and docs/issue-bot.md for the complete security model and operational limits.

WorkflowPromptGuard v0.1.0

WorkflowPromptGuard 0.1.0

Initial public release of the offline policy-as-code scanner for AI-enabled GitHub workflows.

• Traces untrusted GitHub content into write-capable AI agents, secrets, tools, and executable sinks.
• Covers GitHub Agentic Workflows plus Claude, Codex, Copilot, and Gemini action patterns.
• Ships 12 explainable rules, console/JSON/Markdown/SARIF output, a composite GitHub Action, and reasoned/expiring suppressions.
• Includes 37 tests, Python 3.10-3.14 CI, typed APIs, and vulnerable/safe examples.