The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions include the execution of local Python scripts to handle document formatting and structure validation.
Evidence: python3 src/scripts/generate_docx.py is used to export Markdown to Word format.
Evidence: src/scripts/check_sections.py report is used to verify the report structure.
[INDIRECT_PROMPT_INJECTION]: The skill ingests content from external websites, which represents an attack surface for instructions hidden within the crawled data.
Ingestion points: WebSearch and WebFetch are used to collect data from www.gov.cn and various Chinese ministry websites as defined in the 'Default Flow' section of SKILL.md.
Boundary markers: The skill includes specific verification steps ('核实') and mandates that instructions in doc-type-guardrails.md must be followed to prevent fabrication.
Capability inventory: The skill possesses file writing capabilities and the ability to execute local Python scripts based on gathered information.
Sanitization: No explicit technical sanitization of web content is mentioned, although the skill instructs the agent to only record verified facts and avoid fabrication.

ministry-news-daily