The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the WorkorAI marketplace, including job descriptions, candidate profiles, and interview transcripts. This data serves as an ingestion point for potential indirect prompt injection attacks, where malicious instructions could be embedded in the marketplace content to influence the agent's behavior.
Ingestion points: Untrusted data is retrieved using tools such as candidate.search_jobs, employer.search_candidates_for_job, and employer.get_applicant_transcript (SKILL.md).
Boundary markers: The skill advises human review for hiring decisions and requires explicit confirmation for stateful marketplace actions, such as applying to jobs or inviting candidates, which establishes a manual oversight boundary.
Capability inventory: The skill possesses capabilities for stateful actions, including candidate.apply_to_job, employer.invite_candidate, employer.publish_job, and employer.close_job (SKILL.md).
Sanitization: The instructions do not specify any sanitization or filtering of external marketplace content before it is processed by the agent.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for the user to install a remote MCP server from https://workorai.com/mcp and references an external NPM package @workorai/agent-kit. While these are necessary for the skill's primary purpose, they involve downloading and executing code from an external community source.

workorai