The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes tools such as Bash, Read, Glob, and Grep to investigate project state. The instructions in SKILL.md also suggest the use of a piped command (cat .pi/skills-manifest.json | python3 -c "...") to parse project metadata.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted content from the project codebase (e.g., README files, documentation, and source code).
Ingestion points: Accesses project configuration files (pyproject.toml, package.json, etc.), documentation (README.md, CONTEXT.md, docs/), and source code files during the scanning process in assess.py.
Boundary markers: The skill does not implement specific technical delimiters or markers to isolate the content of analyzed files from the agent's instructions.
Capability inventory: Employs file system tools (Bash, Glob, Read, Grep) and executes a Python-based static analysis script (assess.py). It also references the use of external visualization tools via /create-figure.
Sanitization: No evidence of content sanitization or escaping was found for the data read from project files before it is processed by the agent or analysis scripts.

assess