The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local Python scripts (extract_ast.py, git_detective.py, and query_graph.py) and the system git binary to analyze the target repository. Technical review of the Python source code confirms that these commands are executed using subprocess.run with list-based arguments, which is a secure method that prevents shell injection attacks.
[PROMPT_INJECTION]: The skill contains a 'Persistent Instructions' section that directs the agent to store operational rules in shared memory files such as AGENTS.md or CLAUDE.md. While this modifies the agent's long-term behavior guidelines, it is a transparently documented feature intended for session continuity and does not attempt to bypass core safety or ethical constraints.
[DATA_EXFILTRATION]: Although the skill processes local codebase data, it does not include any network-capable modules or external transmission logic. The analysis scripts output data to the local file system (.nexus-map/) or to the standard output for the agent's internal use. Furthermore, the skill's instructions explicitly caution against copying sensitive information like environment variables or secrets.

nexus-mapper