job-hunt-skill
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest job descriptions and titles from untrusted external sources such as LinkedIn and general search engines (identified in SKILL.md Section 3). This creates a vulnerability surface where malicious instructions embedded in remote job postings could influence the agent's behavior during analysis or scoring.
- Ingestion points: SKILL.md Section 3 ('发现公开职位') describes fetching data from LinkedIn job pages, company career sites, and search engine results.
- Boundary markers: The instructions do not specify any delimiters or safety pre-prompts to isolate untrusted JD content from the analysis logic.
- Capability inventory: The skill requires local file writing (SKILL.md Section 6) and browser search capabilities.
- Sanitization: While the skill validates the JavaScript syntax of its output (SKILL.md Section 7), it does not explicitly mention sanitizing or filtering the natural language content extracted from job sites.
- [DYNAMIC_EXECUTION]: The skill generates a self-contained HTML report that includes inline JavaScript and data collections (assets/report-spec.md). The SKILL.md (Section 7) instructs the agent to validate this generated code using
new Function(script). Because the script context incorporates data fetched from untrusted external sources, there is a risk of generating executable content influenced by remote data, though the scope is limited to the local HTML report environment.
Audit Metadata