rjv-gated-build

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill manages external processes by executing shell commands like pkill -9 -f to ensure environment hygiene and terminate orphaned build-related processes during long-running sessions.\n- [EXTERNAL_DOWNLOADS]: The skill utilizes well-known external CLI tools and services, including the OpenAI Codex CLI and local Ollama instances, to delegate mechanical tasks and optimize model usage costs.\n- [PROMPT_INJECTION]: The skill ingests external content from design documents (spec.md) and plan files, creating a potential surface for indirect instructions.\n
  • Ingestion points: Processes implementation requirements from spec.md and maintains durable build memory in .plans/ directory files.\n
  • Boundary markers: Relies on the architecture's "Reviewer" role and its "Grill" phase to adversarialy validate all inputs and implementations against agreed criteria.\n
  • Capability inventory: The skill utilizes file system read/write operations and shell command execution for build management and cleanup.\n
  • Sanitization: Rather than text-based sanitization, the skill employs an independent second agent to verify the safety and correctness of all processed specifications and code changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 04:59 AM
Security Audit — agent-trust-hub — rjv-gated-build