The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/search.py uses subprocess.run to execute a local sibling script (web-fetch/scripts/fetch.py) via the uv tool. It passes URL data retrieved from ministry metadata as command-line arguments.
[COMMAND_EXECUTION]: The SKILL.md file provides instructions for the agent to execute a shell command (find / -name departments.py ...) to locate the skill's base directory on the host system if the environment path is not predefined.
[EXTERNAL_DOWNLOADS]: The skill initiates network requests through its helper script to fetch policy documents and metadata from multiple official Chinese government domains, including ndrc.gov.cn, miit.gov.cn, mem.gov.cn, and the State Council's policy library.
[PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it ingests and interprets external HTML and PDF content from policy sites without sanitizing the text for embedded instructions.
Ingestion points: External policy text and attachment contents fetched in SKILL.md Step 3 and processed by scripts/search.py.
Boundary markers: None. The skill does not define specific delimiters or instructions to ignore potential commands embedded within the fetched policy documents.
Capability inventory: The skill has the ability to execute local shell commands and Python scripts via subprocess and perform network operations.
Sanitization: Employs basic HTML tag stripping (strip_tags) in scripts/search.py, but lacks filtering for natural language instructions in the final text provided to the agent for interpretation.

policy-interpretation