The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes job advertisements from Jobnet.dk, representing a surface for untrusted third-party data ingestion. Ingestion points: Data is fetched via API responses in cli/src/commands/search.ts and cli/src/commands/detail.ts. Boundary markers: Absent. The CLI does not wrap its output in specific delimiters or warnings for the agent. Capability inventory: The skill can execute its local CLI via Bash and perform network requests to the Jobnet portal (identified in cli/src/helpers.ts). Sanitization: Present and robust. The stripInvisible function in cli/src/helpers.ts removes control characters, bidi overrides, and Unicode tag sequences (U+E0000-E007F). Additionally, visibleHtml and stripHtml functions strip non-rendering markup, comments, and scripts from job descriptions before they reach the agent context.
[SAFE]: The skill interacts with the official Danish job portal, Jobnet.dk, which is a well-known and authoritative national service. This activity is consistent with the skill's primary purpose.
[SAFE]: Static detection of steganographic patterns in cli/tests/hidden-text.test.ts is attributed to a security test suite. This suite specifically verifies that the skill's defensive logic effectively blocks and removes prompt injection payloads from processed data.

jobnet-search