py-shiny-release
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill provides detailed instructions for the agent to execute shell commands necessary for a software release cycle. This includes repository management (
git checkout,git tag,git push), CI monitoring and release creation via the GitHub CLI (gh pr checks,gh run list,gh release create), and cleanup of development servers usinglsofandkill. These operations are standard for release engineering and include explicit user confirmation gates. - [EXTERNAL_DOWNLOADS]: The skill interacts with external registries and APIs to manage dependencies and verify release state. It fetches metadata from PyPI and GitHub APIs using
curlandgh api, and installs development tools like Playwright and SCSS compilers viapipandnpm. All target resources are either well-known public registries or official repositories within the author's organization. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes data from external sources.
- Ingestion points: The agent reads external content including changelogs, Git commit histories, GitHub PR comments, and JSON responses from GitHub and PyPI APIs.
- Boundary markers: The instructions do not specify explicit delimiters to isolate untrusted data when presented to the model's context.
- Capability inventory: The skill possesses capabilities to execute shell commands, install packages, and modify files within the local repository environment.
- Sanitization: The skill uses structured tools like
jqandgrepto extract specific information from external data, reducing the likelihood of processing hidden instructions, although raw text from changelogs is ultimately reviewed by the agent. - [DYNAMIC_EXECUTION]: The skill includes a helper script (
references/test_shinylive_site.py) that usessubprocess.runto call the GitHub CLI. Analysis shows this is a safe implementation using a static list of arguments to fetch example indices, serving as a legitimate testing utility for the release process.
Audit Metadata