The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local git commands via scripts/git_detective.py to analyze repository history. This is implemented using standard practices (passing arguments as a list to subprocess.run) and is necessary for its intended forensic functionality. \n- [EXTERNAL_DOWNLOADS]: The skill depends on well-known and reputable third-party libraries for code parsing and file filtering, specifically tree-sitter, tree-sitter-language-pack, and pathspec, as documented in scripts/requirements.txt. \n- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it ingests and processes untrusted content from the repository being analyzed. \n
Ingestion points: scripts/extract_ast.py and scripts/git_detective.py read all source files and Git metadata within the provided repository path. \n
Boundary markers: Analysis outputs written to the .nexus-map/ directory do not utilize explicit structural delimiters or "ignore" instructions to isolate untrusted content from the agent's operating context. \n
Capability inventory: The skill possesses the ability to read all files in the project, write to the file system, and execute its own set of Python scripts. \n
Sanitization: The skill provides explicit directives to the agent in SKILL.md (守则5) to protect sensitive information like .env files and to avoid executing any build or test scripts found within the target repository.

nexus-mapper