binlee-clinic-operations

Pass

Audited by Gen Agent Trust Hub on Jul 16, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONNO_CODE
Full Analysis
  • [DATA_EXFILTRATION]: The skill reads from a parent directory path '../binlee-source-library/references/evidence-policy.md' as instructed in SKILL.md. While this appears to be a vendor-managed resource library, accessing files outside the skill's own directory is a data exposure risk.
  • [PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it ingests data from a local medical beauty corpus to generate diagnoses. Evidence Chain: 1. Ingestion points: SKILL.md specifies reading 'evidence-policy.md' and searching through external corpus files. 2. Boundary markers: none are used in SKILL.md to isolate ingested data from agent instructions. 3. Capability inventory: all instructions in SKILL.md are limited to text generation; no high-risk capabilities like file-writing or network requests were identified. 4. Sanitization: no sanitization or filtering logic is mentioned in SKILL.md for the processed corpus data.
  • [NO_CODE]: This skill consists entirely of configuration and instructional markdown files and does not include any executable scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 16, 2026, 01:23 AM
Security Audit — agent-trust-hub — binlee-clinic-operations