The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various system commands and local scripts as part of the release workflow.
Evidence: Executes gh auth status, git fetch, uv version, npm install, and a local release script located at desktop/scripts/release.sh.
[EXTERNAL_DOWNLOADS]: The skill downloads release metadata and checks the status of artifacts hosted on GitHub.
Evidence: Fetches latest.json from https://github.com/Shpigford/nurb/releases/download/desktop-latest/latest.json to verify the update feed.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub pull requests which could theoretically influence agent behavior.
Ingestion points: Reads pull request titles and metadata using gh pr list --json title,mergedAt (found in SKILL.md).
Boundary markers: No explicit delimiters or instructions are used to isolate pull request data from the agent's core instructions.
Capability inventory: The skill has extensive permissions including file system writes, git commits/pushes, and shell script execution (scripts/release.sh).
Sanitization: No sanitization logic is present for the pull request data before it is processed by the agent to generate changelogs.

release