The Agent Skills Directory

[SAFE]: The skill uses official tools (GitHub CLI) and standard system utilities (awk, bash, date) for its operations. All network interactions are directed at GitHub's official APIs.
[PROMPT_INJECTION]: The skill processes untrusted content from external sources (GitHub repository descriptions and READMEs), which is a surface for indirect prompt injection. This is an inherent risk for research-oriented skills. The provided scripts perform basic sanitization by removing control characters (tabs, newlines) from repository descriptions before processing.
Ingestion points: README content and repository descriptions fetched via gh api and gh search in scripts/gh-search.sh and references/evaluate-code.md.
Boundary markers: The instructions advise the agent to classify candidates into a structured table before deep reading, which provides an internal context boundary.
Capability inventory: File read/write (via redirection), shell execution (gh, awk), and network access (via gh).
Sanitization: scripts/gh-search.sh uses awk to remove tabs and newlines from repository descriptions.
[SAFE]: No hardcoded credentials, malicious persistence mechanisms, or obfuscated code patterns were found. The use of base64 -d in the documentation is for decoding the standard Base64-encoded file content returned by the GitHub API.

run-github-scout