The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands and Python code to resolve paths and read documentation. Evidence: python3 -c "from qwenpaw.constant import DOCS_DIR...", which qwenpaw, find ~/.qwenpaw/memory/, and cat <doc_path> are used to locate and read files.
[PROMPT_INJECTION]: The skill processes untrusted local data, creating an indirect prompt injection surface. (1) Ingestion points: Documentation content is read from the local file system (SKILL.md, Step 3). (2) Boundary markers: Absent; no instructions wrap ingested content in delimiters. (3) Capability inventory: The agent has shell execution and file system access. (4) Sanitization: Absent; no validation is performed on documentation content.

guidance