The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted source code from the local repository to build structural AST maps. This data is then presented to the agent, which could be exploited by malicious code identifiers or comments designed to influence agent behavior.
Ingestion points: Source code files are read by scripts/extract_ast.py using Path.read_bytes().
Boundary markers: None. Extracted structural metadata is passed directly to the agent without specific delimiters or isolation instructions.
Capability inventory: The skill involves reading files, writing JSON metadata, and executing shell commands through scripts/git_detective.py.
Sanitization: No explicit sanitization or filtering of extracted strings (like class or function names) is performed before providing them to the agent context.
[COMMAND_EXECUTION]: The skill utilizes system commands to analyze repository history.
Evidence: scripts/git_detective.py contains calls to subprocess.run for executing git commands (e.g., git log) to identify hotspots and logical coupling between files.

nexus-query