The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from job listings fetched via the jobbank.dk RSS feed and HTML pages.
Ingestion points: The search command in cli/src/commands/search.ts (RSS item fields) and the detail command in cli/src/commands/detail.ts (HTML body and JSON-LD blocks).
Boundary markers: The CLI does not explicitly provide boundary markers (like "the following is untrusted content") in its output to the agent.
Capability inventory: The skill is limited to reading data from a specific external domain and outputting text to stdout; it has no file system write access or command execution capabilities.
Sanitization: The detail command includes a sanitization step that strips HTML tags from the description before returning it to the agent, reducing the risk of content-based injection.
[SAFE]: The CLI tool implements secure network practices, including a 15-second timeout via AbortSignal and an exponential backoff strategy for handling 429 and 5xx errors, preventing potential denial-of-service conditions or hangs.

jobbank-search