The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Automated security scans identified commands piping curl output to python3. These are verified as safe data-parsing operations using a local Python script (-c) to extract version numbers from trusted project metadata on GitHub and PyPI. No remote code is downloaded or executed.
[COMMAND_EXECUTION]: The skill executes a variety of local development tools and scripts, including uv, npm, pytest, and the project's own scripts/release.sh. It also performs environment checks for sensitive local files such as the Tauri signing key (~/.tauri/nurb-desktop.key) to ensure the system is ready for a production release.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the titles and descriptions of merged Pull Requests via gh pr list to construct a changelog.
Ingestion points: Pull Request metadata retrieved from the GitHub API in SKILL.md.
Boundary markers: None present; the skill directly interpolates external PR content into its context.
Capability inventory: The skill has significant capabilities including writing to the file system, committing to git, pushing to remote branches, and executing build scripts.
Sanitization: There are no explicit instructions to sanitize or filter the content of the Pull Requests before they are processed by the agent.

release