openspec-onboard
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection Surface: The skill performs codebase analysis by reading local files to identify improvement opportunities like
TODOcomments or code patterns. This ingestion of untrusted data represents a surface where hidden instructions in the codebase could potentially influence the agent's behavior during task creation or implementation phases. - Command Execution: The skill executes shell commands using the
npx openspecutility to interact with the project's change management system. While these are intended operations for the tool, they involve subprocess execution and interaction with external package registries.
Audit Metadata