workflow-big-feature
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill orchestrates a complex workflow that ingests untrusted data from external sources via research tools, which then informs critical design and implementation steps.
- Ingestion points: The
$web-research --chain=deep-diveand$market-analysissteps fetch and process external data from the internet. - Capability inventory: The workflow includes powerful capabilities such as spawning sub-agents (
spawn_agent), generating code ($plan --mode=execute), and performing automated testing. - Boundary markers: There are no explicit instructions within this file for the use of delimiters or 'ignore embedded instructions' warnings when processing the results of external research.
- Sanitization: The instructions do not mention specific sanitization or validation logic for external content before it is interpolated into the design documents or implementation plans.
- [COMMAND_EXECUTION]: The skill contains instructions for executing a local Node.js script to manage its own configuration and documentation.
- Evidence: The instruction
node .claude/scripts/lib/workflow-skill-contract.cjs --writetriggers the execution of a script within the project's own repository structure. - [DYNAMIC_EXECUTION]: The workflow is designed to generate code and tests dynamically based on the research and specification phases.
- Evidence: The
$plan --mode=executestep is explicitly tasked with writing phase code and its corresponding tests, which are then executed during the build and verification stages.
Audit Metadata