The Agent Skills Directory

[PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection (Category 8) because it ingests and processes untrusted codebase files. Ingestion points: Files are read from a user-provided repository path by scripts/analyze.py. Boundary markers: The instructions mandate reading a structured JSON report and citing source files, which provides some structure but no hard isolation. Capability inventory: The skill has access to Bash(python3:*), Bash(pip:*), Bash(git:*), and Read (defined in SKILL.md). Sanitization: There is no explicit sanitization of symbol names or file content extracted from the codebase before they are presented to the model.
[COMMAND_EXECUTION]: The scripts/analyze.py script executes several external commands including git log, tokei, and scc via subprocess.run to gather repository insights. These operations are performed with the agent's system permissions.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of external Python libraries specified in scripts/requirements.txt, including tree-sitter, pathspec, networkx, and tomli.

codebase-onboarding