orchestrate-implementation

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes external, potentially untrusted project documentation to direct automated agents.
  • Ingestion points: The skill reads ADRs, specifications, issues, and plans during the 'Intake and preflight' phase (SKILL.md).
  • Boundary markers: The instructions emphasize extracting constraints and acceptance criteria but do not define explicit delimiters or 'ignore' instructions for the processed content.
  • Capability inventory: The orchestrator dispatches subagents with mutation authority in managed worktrees (references/pi-dispatch.md).
  • Sanitization: No explicit sanitization or filtering of the intake data is described; the skill relies on scout agents to normalize content.
  • [SAFE]: The skill enforces significant security best practices, including repository cleanliness verification, mandatory git isolation for parallel mutation, and strict human/supervisor gates for high-risk operations like pushing code, merging, or deployment. The use of subagents is scoped by specific task contracts and bounded file references.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 09:06 PM
Security Audit — agent-trust-hub — orchestrate-implementation