The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches content from external URLs using services like r.jina.ai and defuddle.md. These are established and well-known services for web-to-markdown conversion and are used here for their intended purpose.
[COMMAND_EXECUTION]: The skill employs local CLI tools such as curl, npx, pdftotext, and marker_single to perform data fetching and PDF extraction. These operations are scoped to processing the content requested by the user and do not involve privilege escalation or unauthorized system access.
[DATA_EXPOSURE]: The skill facilitates the download of web content and images to the local ~/Downloads/ directory. This is standard behavior for a reading utility and is triggered only by explicit user requests or specific workflow requirements.
[INDIRECT_PROMPT_INJECTION]: As the skill processes untrusted data from the internet, it implements safeguards by instructing the agent to focus solely on conversion and storage without interpreting or analyzing the content.
Ingestion points: Content is ingested from user-provided URLs and PDF files in SKILL.md.
Boundary markers: Explicit instructions in SKILL.md tell the agent: 'No analysis, no summary, no discussion of the content unless explicitly asked' and 'Your job is conversion and storage, not interpretation.'
Capability inventory: The skill uses scripts/fetch.sh (curl), scripts/fetch_feishu.py (requests), and scripts/fetch_weixin.py (playwright) for network operations, and writes files to ~/Downloads/.
Sanitization: Fetched content is converted to Markdown and truncated to 200 lines to limit the influence of large payloads.

read