hr-resume-screener
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted external data (resumes and cover letters), which creates a surface for indirect prompt injection attacks where hidden instructions in applicant documents could attempt to manipulate the AI's evaluation scores or logic.
- Ingestion points: The skill accepts 'resume' (PDF/Text) and ' 자기소개서' (Cover Letter) as inputs, as described in SKILL.md.
- Boundary markers: It uses mandatory output headers (human-review warnings) and structured audit trails to define output boundaries, though it lacks explicit technical delimiters for the untrusted input text itself.
- Capability inventory: The skill performs text analysis, calculates SHA256 hashes of inputs, and references the 'moai-recruiter:hr-employment' skill for broader recruiter tasks.
- Sanitization: It implements automated masking of protected attributes (PII such as photos, birthdates, and family information) to adhere to hiring laws, but it does not provide specific mechanisms to filter out command-injection strings within the text.
- [SAFE]: The skill is strictly limited to per-candidate evaluation and explicitly prohibits automated decision-making (auto-rejection or auto-acceptance), which reduces the impact of potential AI bias or manipulation.
- [SAFE]: All external references point to legitimate legal, regulatory, and professional HR resources (e.g., NCS, EEOC, Greenhouse, and legal blogs). All internal skill references align with the vendor's known resource patterns.
Audit Metadata