The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill displays external data from the GitHub Traffic API which can contain user-controlled content like HTTP referrers and repository page paths. If an AI agent interprets this output, it could be exposed to indirect prompt injection.
Ingestion points: repos/{repo}/traffic/popular/referrers and repos/{repo}/traffic/popular/paths endpoints in scripts/github_traffic.py.
Boundary markers: Absent; data values are rendered directly into plain-text tables without encapsulation or specific instruction delimiters.
Capability inventory: Structured file writing to ~/.github-traffic/ and process spawning via subprocess.run.
Sanitization: None; fields like referrer and path are displayed directly to standard output as strings.
[COMMAND_EXECUTION]: The script executes external commands using python's subprocess module to query the GitHub CLI.
Evidence: Spawns gh api inside the gh_api function in scripts/github_traffic.py. This is handled safely via argument arrays without enabling shell execution parsing (shell=True).

github-traffic