The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. Malicious instructions embedded in Knowledge Planet posts or PDF files could potentially influence the agent's behavior.\n
Ingestion points: The skill fetches post content and PDF text from api.zsxq.com via fetch_topics.js.\n
Boundary markers: No delimiters or safety instructions are used when interpolating external content into the AI prompt in SKILL.md.\n
Capability inventory: The skill executes local commands and makes network requests to fetch data.\n
Sanitization: The skill does not sanitize or filter the content retrieved from Knowledge Planet before passing it to the model.\n- [COMMAND_EXECUTION]: The skill performs command execution to run its data fetching logic.\n
Evidence: SKILL.md instructs the agent to run node fetch_topics.js and bash install.sh.\n- [EXTERNAL_DOWNLOADS]: The skill downloads external data to fulfill its primary purpose.\n
Evidence: fetch_topics.js fetches JSON data and PDF files from api.zsxq.com. As this is a well-known service related to the skill's purpose, this is documented neutrally.

zsxq-summary