The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/miner.py is programmed to read GitHub authentication tokens directly from a hardcoded sensitive path in the user's home directory (~/.github-token).
[COMMAND_EXECUTION]: The validation script scripts/validate.py uses the __import__() function for dynamic module loading to verify the presence of the github library at runtime.
[PROMPT_INJECTION]: The skill processes untrusted data from GitHub PR comments, representing an indirect prompt injection surface.
Ingestion points: Fetches PR review comments from external repositories via the GitHub API in scripts/miner.py.
Boundary markers: Absent; the output JSON does not contain markers or instructions for downstream agents to isolate or ignore embedded content.
Capability inventory: The skill includes access to Bash, Write, Edit, and Task tools across its lifecycle.
Sanitization: No sanitization or filtering of external comment body text is performed during the extraction process.
[DATA_EXFILTRATION]: The skill performs automated extraction of repository data and PR history from GitHub and writes this information to local files, constituting a flow of potentially sensitive project metadata into the local environment.

pr-miner