301-spec-implementation

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill defines a disciplined orchestrator-worker model for software implementation, separating concerns between design, test-writing, implementation, and validation. It adheres to standard engineering practices like tests-first development and phase-isolated context handling.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied specifications and change requests to drive automated coding and validation actions.\n
  • Ingestion points: Feature files, change requests, and design documentation referenced in SKILL.md (Phase 0, Step 1, 2, 4).\n
  • Boundary markers: SKILL.md defines 'Task Packets' (Operating Model point 6) to scope subagent context, though explicit instruction-ignoring delimiters are absent.\n
  • Capability inventory: Subagents @coder (Phase 1, 2, 3, 4) and @validator (Step 6, 8, 10, 12, 14, 15) perform file writes and shell command execution in SKILL.md.\n
  • Sanitization: Absent in SKILL.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 06:51 PM
Security Audit — agent-trust-hub — 301-spec-implementation