The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands via a Bash function to interact with the GitHub CLI (gh), git, and npm. These operations are central to its purpose of automating repository tasks such as cloning, dependency updates, and pull request creation.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from external GitHub repositories, including package.json and documentation files, creating a potential attack surface where malicious content in those repositories could influence agent behavior.
Ingestion points: Reads repository metadata, file contents (e.g., package.json, CLAUDE.md), and search results from the GitHub API.
Boundary markers: None explicitly implemented in the provided examples to isolate external content from instructions.
Capability inventory: Extensive command execution capabilities including npm test, git push, and gh api calls.
Sanitization: Relies on standard shell utilities and jq for parsing, with no specific prompt-level sanitization for ingested text.
[SAFE]: The skill uses base64 encoding and decoding solely for interacting with the GitHub API's file content requirements, which is standard practice and not an attempt at obfuscation. It also correctly handles sensitive configurations by referencing environment variables (e.g., process.env.WEBHOOK_SECRET).

github-multi-repo