The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and parses external content from Knowledge Planet (ZSXQ) article and topic URLs, which introduces an indirect prompt injection surface where instructions embedded in the external content could attempt to influence agent behavior.
Ingestion points: The fetch_source function in scripts/tuwen_likes_export.py retrieves content from user-provided URLs using an external CLI tool.
Boundary markers: The skill does not implement specific technical boundary markers for the data ingestion phase, but it enforces a mandatory 'human-in-the-loop' confirmation step requiring the user to review the extracted topic list before statistical processing occurs.
Capability inventory: The script has the capability to write files (CSV/JSON/Markdown) to the local filesystem and execute shell commands via subprocess calls.
Sanitization: Content is processed using HTMLParser and regular expressions to extract specific structured data, reducing the likelihood of raw instruction execution.
[COMMAND_EXECUTION]: The Python script scripts/tuwen_likes_export.py utilizes subprocess.run to interact with the local system and external utilities.
Evidence: The run_json function executes zsxq-cli to perform API calls, and the resolve_topic_link function uses curl to resolve URL redirects. While these are executed using a list format (reducing shell injection risk), they represent execution of external binaries based on inputs derived from processed URLs.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch data from zsxq.com domains and uses curl to resolve short links.
Evidence: The script fetches article and topic details via zsxq-cli and resolves redirection for t.zsxq.com links to identify target topic IDs.

zsxq-tuwen-likes-export