The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script connects to the official GitHub API (api.github.com) to retrieve repository metadata, search results, and release information. These are calls to a well-known service for its intended purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and prints untrusted data from GitHub, such as repository descriptions, release notes, and topics. This content can be controlled by third-party repository owners to include instructions aimed at an agent processing the output.
Ingestion points: Data is fetched from GitHub in scripts/gh_trending.py via the trending, stars, repo, releases, and search commands.
Boundary markers: No delimiters or safety warnings are used to separate untrusted external content from the agent's instructions.
Capability inventory: The skill is limited to making network GET requests to GitHub and printing results; it does not perform sensitive file operations or execute shell commands based on the fetched data.
Sanitization: The script performs basic truncation of descriptions and release notes to 120 characters, which limits the volume of injected text but does not filter for malicious instructions.

ez-github