resume-tailor-ingest
Pass
Audited by Gen Agent Trust Hub on Aug 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a formatting and cleaning phase that explicitly removes zero-width characters (U+200B, U+FEFF, etc.) from user-provided resumes, which serves as a defense against common text-based obfuscation techniques.\n- [SAFE]: The skill handles third-party data obfuscation by decoding Private Use Area (PUA) characters (U+E031–U+E03A) used by specific recruitment platforms into standard numeric digits, ensuring data integrity during the import process.\n- [COMMAND_EXECUTION]: The skill executes local scripts via Node.js, such as scripts/import_jobs.mjs and scripts/import_screenshot_job.mjs, to process job materials and screenshots. These operations are limited to the skill's local directory structure.\n- [DATA_EXFILTRATION]: The skill reads resume and job data to construct candidate profiles and job JSON files in the output directory. This behavior is consistent with the skill's primary function and does not involve external network exfiltration.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it ingests and processes untrusted content from resumes and job descriptions.\n
- Ingestion points: Processes text, document files, and screenshots from recruitment platforms as described in SKILL.md and references/screenshot-intake.md.\n
- Boundary markers: Uses structured output formats (JSON and STAR-based Markdown) for extraction, but does not explicitly define delimiters for untrusted input within the LLM prompt context.\n
- Capability inventory: Executes local Node.js scripts to process the ingested data.\n
- Sanitization: Includes a cleaning step to normalize formatting and remove hidden characters from inputs.
Audit Metadata