The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/check_discoverability.py executes the GitHub CLI (gh) via subprocess.run. The implementation is secure as it uses a list of arguments rather than a shell string, effectively preventing shell injection. The script's operations are restricted to the repository specified by the agent or user, and it leverages data sourced directly from GitHub's official API responses.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from GitHub repositories, creating a potential surface for indirect prompt injection.
Ingestion points: The scripts/check_discoverability.py script fetches repositoryTopics and homepageUrl from external repositories and prints them for the agent to review.
Boundary markers: The skill does not use specific delimiters or explicit instructions to the agent to disregard instructions potentially embedded within the repository metadata.
Capability inventory: The agent is empowered to use gh repo edit to modify repository settings (description, topics, homepage) based on its interpretation of the audit results.
Sanitization: The skill does not perform sanitization, filtering, or validation of the retrieved metadata strings before presenting them to the agent.

repo-discoverability