writing-technical-plans

Pass

Audited by Gen Agent Trust Hub on Aug 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute Node.js scripts (scripts/lint-plan.mjs and scripts/audit-remap.mjs) via the shell to validate the structure, style, and content integrity of technical plans. These scripts perform local file system operations to read documents and generate JSON audit reports.\n- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests, revises, and remaps external markdown documents. A malicious input file could attempt to embed instructions to influence the agent's behavior during the reformatting process.\n
  • Ingestion points: External markdown files processed during the 'Revise' and 'Lossless structural remap' modes as defined in SKILL.md.\n
  • Boundary markers: Structural integrity is enforced through a manifest.json and unit-based SHA-256 hashing within the scripts/audit-remap.mjs script.\n
  • Capability inventory: The skill environment permits reading and writing local files and executing shell commands to run Node.js scripts.\n
  • Sanitization: The audit-remap.mjs script explicitly identifies and rejects HTML comments and raw HTML in certified targets to prevent the concealment of unauthorized instructions or data.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 5, 2026, 05:04 PM
Security Audit — agent-trust-hub — writing-technical-plans