sherpa

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it is designed to ingest and process untrusted user-supplied epics, stories, and task descriptions for decomposition.
  • Ingestion points: The skill takes 'Raw MAP input' from users or external AI-generated task lists (like those from Linear Agent) as described in SKILL.md and reference/task-breakdown.md.
  • Boundary markers: It employs structured handoff formats, such as ## NEXUS_HANDOFF and ## SHERPA_TO_IMPL_HANDOFF, to attempt context isolation when delegating to other agents.
  • Capability inventory: The skill coordinates several high-privilege agents including Builder (code implementation) and Forge (prototyping), creating a path where injected instructions in a task description could lead to unauthorized code changes.
  • Sanitization: The instructions implement a 'Quality Gate' using the INVEST checklist and mandatory exit criteria to validate task integrity, though these are primarily behavioral rather than technical sanitization filters.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 06:55 AM
Security Audit — agent-trust-hub — sherpa