concept-design-cn

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides several Python utility scripts (scripts/concept_lint.py, scripts/concept_map.py, and scripts/spec_syntax.py) designed to be executed locally by the agent. These scripts perform structural validation of design documents and generate Mermaid diagrams. They use only Python standard library modules and include safety features such as atomic file replacement and symbolic link protection.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided specification files (CONCEPT.md and SYNCS.md) to guide software implementation and review tasks. This creates an attack surface for indirect prompt injection if those files contain malicious instructions.
  • Ingestion points: Specification files and project source code defined in the project root relative to the agent's environment.
  • Boundary markers: The skill uses strict Markdown headers (e.g., # concept, ## actions) and structural syntax to delimit instructions.
  • Capability inventory: The agent can read/write files and generate code based on the specifications. It can also execute the included Python scripts.
  • Sanitization: The skill includes a structural linter (scripts/concept_lint.py) that the agent is instructed to run to validate that input files conform to the expected schema, which helps mitigate accidental obedience to malformed instructions.
  • [SAFE]: Analysis of all 27 files confirms that the skill strictly follows software design principles. There are no external network calls, hardcoded credentials, or obfuscated code blocks. The provided scripts are transparent and limited in scope to the skill's stated purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 02:48 AM