The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill integrates with GitHub PR comments, which are untrusted external sources.
Ingestion points: references/github-setup.md identifies github-pr-comment as a data source for extracting URLs and SHAs using regex.
Boundary markers: Explicit warning instructions in references/github-setup.md advise that PR comments are untrusted data and should not be followed or copied blindly.
Capability inventory: The skill uses the gauge CLI to run agent sessions and execute tests as described in SKILL.md.
Sanitization: Employs regex pattern matching to extract specific fields (URL and SHA) rather than consuming the entire comment body.
[DYNAMIC_EXECUTION]: The tool is designed to download and execute binaries for Native CLI evaluation.
Evidence: references/github-setup.md describes how the Gauge CLI downloads Linux x86-64 ELF executables from GitHub Actions artifacts, restores execution permissions, and adds them to the system PATH.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the @withgauge/cli package from the npm registry.
Evidence: SKILL.md instructs the user to run npm install -g @withgauge/cli to install the necessary command-line tool.

gauge-agents