linkedin-search
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches job data from LinkedIn's public guest endpoints. This is normal behavior for its primary purpose and targets a well-known service.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and parses HTML content from external LinkedIn pages, which could theoretically contain malicious instructions designed to influence the AI agent processing the output.
- Ingestion points: Fetches job descriptions and metadata from LinkedIn in
cli/src/helpers.ts. - Boundary markers: Absent; the output is returned as raw text or structured JSON without specific delimiters or warnings for the agent to ignore embedded instructions.
- Capability inventory: The skill performs network requests to LinkedIn and executes its CLI tool via
bun run. - Sanitization: Performs basic HTML tag stripping and entity decoding in
cli/src/helpers.tsto clean the content. - [COMMAND_EXECUTION]: The skill executes its own TypeScript CLI logic using the
Bashtool andbun. This execution is restricted to the skill's own source files.
Audit Metadata