break-req

Pass

Audited by Gen Agent Trust Hub on May 11, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The 'Skill Isolation' section contains instructions that attempt to override the agent's standard behavior and platform-level orchestration. It commands the agent to treat the skill as the 'sole active authority' and to 'suppress' any other skills or external interruptions during its execution.
  • [PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by processing user-provided requirement data without safety boundaries while possessing the capability to explore the local codebase.
  • Ingestion points: User-supplied implementation requirements during the 'interrogation' workflow in SKILL.md.
  • Boundary markers: Absent; no instructions are provided to delimit user data or ignore instructions embedded within the requirements.
  • Capability inventory: Read-only exploration of the local codebase (referenced in SKILL.md).
  • Sanitization: Absent; the skill does not specify any validation or filtering of the ingested technical requirements.
Audit Metadata
Risk Level
SAFE
Analyzed
May 11, 2026, 05:58 PM