job-search-agent
Warn
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill manages system-level persistence by configuring automated tasks using
cronandlaunchd. It instructs the agent to compose and register run recipes that execute periodically without an active user session to ensure search reliability. - [PROMPT_INJECTION]: The system is subject to indirect prompt injection risks because it ingests and analyzes job postings from third-party websites (LinkedIn, Ashby, Greenhouse, Lever). A malicious job description could contain hidden instructions designed to manipulate the evaluation logic or influence the agent's actions.
- Ingestion points: Job postings fetched from external sources via the agent-data CLI.
- Boundary markers: No specific boundary markers or 'ignore' instructions are defined for the evaluation of external content.
- Capability inventory: Local file system writes to
~/.job-search/, system scheduler modifications, and network requests via vendor-provided tools. - Sanitization: The instructions do not specify any sanitization or filtering of the external job description text before it is analyzed by a subagent.
- [DATA_EXFILTRATION]: The skill performs network operations to fetch data from LinkedIn, Ashby, Greenhouse, and Lever, and communicates with the vendor's service at
agent-data.motie.dev. While these are vendor-managed resources, they involve external data flows and API quota management.
Audit Metadata