The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted third-party job listings from Jobindex.dk. It implements a dedicated security layer to prevent indirect prompt injection. \n
Ingestion points: Fetches job search results and details from www.jobindex.dk in cli/src/commands/search.ts and cli/src/commands/detail.ts.\n
Boundary markers: The skill does not use specific boundary delimiters but relies on comprehensive sanitization of the input data to remove potential attack vectors.\n
Capability inventory: The skill uses Bash to run a local CLI which performs network fetches via fetch and outputs processed job data. It does not write to the file system or execute arbitrary shell commands with data content.\n
Sanitization: The stripInvisible function in cli/src/helpers.ts uses a regex to remove Unicode tag characters (U+E0000-E007F), zero-width characters, and control codes. The visibleHtml function in cli/src/helpers.ts strips non-rendered HTML tags and elements marked as hidden to ensure only human-visible text is processed.\n- [COMMAND_EXECUTION]: The skill uses a local CLI tool (cli/src/cli.ts) to interact with Jobindex.dk. It performs thorough manual validation of all command-line flags and positional arguments before execution to prevent command injection, as evidenced in cli/src/cli.ts and verified by the test suite.\n- [EXTERNAL_DOWNLOADS]: The skill fetches HTML content from https://www.jobindex.dk. The buildUrl function in cli/src/commands/detail.ts enforces strict validation, only allowing requests to the jobindex.dk domain or its subdomains, preventing arbitrary URL fetching or SSRF-like behavior.

jobindex-search