hiring

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill maintains state by reading and writing to local context files representing manager and employee profiles.
  • Ingestion points: The skill reads from .agents/em-context.md and .agents/reports/[name].md (found in SKILL.md).
  • Boundary markers: There are no explicit delimiters used when reading these files to distinguish stored data from instructions.
  • Capability inventory: The skill's capabilities are limited to reading and writing these specific local files; it contains no network operations or arbitrary command execution.
  • Sanitization: The instructions include a heuristic for the agent to only save "stable facts and patterns" rather than "guesses" or "unresolved interpretations," which acts as a basic instruction-level filter.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 05:46 AM
Security Audit — agent-trust-hub — hiring