The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes job descriptions from an external source (Jobnet.dk), which contains attacker-controllable content that could theoretically include instructions meant to influence the agent's behavior.
Ingestion points: The detail command in cli/src/commands/detail.ts fetches raw job description content from the /FindJob/JobAdDetails/ API endpoint.
Boundary markers: The agent instructions in SKILL.md do not currently define explicit delimiters or instructions to ignore potential commands embedded within the job description text.
Capability inventory: The skill is granted Bash tool access to run a local CLI which performs network requests to the jobnet.dk domain.
Sanitization: The CLI includes a stripHtml helper in cli/src/helpers.ts that removes HTML tags and decodes common entities, which significantly reduces the risk of hidden payload execution within the agent's context.
[COMMAND_EXECUTION]: The skill exposes a command-line interface (cli/src/cli.ts) executed via the Bash tool to interact with the Jobnet API.
The tool utilizes the Zod library to strictly validate all user-supplied flags (such as page, radius, and limit), ensuring they are valid integers and within permitted ranges, which prevents common command injection attacks.
[EXTERNAL_DOWNLOADS]: The CLI fetches job listing and occupation data from the official Danish government job portal API.
The requests are directed to the legitimate https://jobnet.dk/bff endpoints, which is the documented and intended purpose of the skill.

jobnet-search