execute-blueprint
Pass
Audited by Gen Agent Trust Hub on May 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions and subagent prompts exhibit no signs of malicious intent, obfuscation, or unauthorized data exfiltration. The inclusion of hard gates for user configuration and adversarial review requirements is a strong security practice.\n- [COMMAND_EXECUTION]: The verification subagent runs shell commands for testing, linting, and type-checking based on the toolchain defined in the project's configuration files or implementation plans. This is a core part of the build-review-verify cycle.\n- [PROMPT_INJECTION]: The skill processes markdown implementation plans as a source of truth for subagent instructions, which presents a surface for indirect prompt injection.\n
- Ingestion points: Implementation plans and milestone documents found in the
docs/plans/directory.\n - Boundary markers: Instructions are passed to subagents via template placeholders without explicit boundary delimiters.\n
- Capability inventory: Subagents can perform file operations and execute shell commands through the Agent tool.\n
- Sanitization: The workflow relies on the review subagent's security analysis and the serial verification process to validate instructions and code changes.
Audit Metadata