The Agent Skills Directory

[COMMAND_EXECUTION]: Extensive use of shell commands and subprocesses to coordinate complex workflows across multiple projects.\n
Evidence: The run.sh script executes find, grep, and coordinates several other skills (project-state, cleanup, ingest-code, etc.) via their respective entry points.\n
Evidence: autofix_docstrings.py uses subprocess.run to call treesitter and scillm for analysis and docstring generation.\n- [COMMAND_EXECUTION]: Employs inline Python scripts for configuration parsing and data manipulation.\n
Evidence: run.sh uses python3 -c to parse ~/.agent-inbox/projects.json and project-specific .monitor-codebase.json files.\n- [PROMPT_INJECTION]: Potential surface for indirect prompt injection from untrusted source code being analyzed.\n
Ingestion points: The skill reads local project source code (.py, .tsx, etc.) and configuration files.\n
Boundary markers: Employs JSON formatting for violation data in generated task files for dispatched agents.\n
Capability inventory: The skill can modify local source files, execute external scripts, and register scheduled tasks.\n
Sanitization: Includes truncation of content sent to language models and stripping of formatting characters from model responses in autofix_docstrings.py.

monitor-codebase