openspec-onboard

Pass

Audited by Gen Agent Trust Hub on Apr 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes various local shell commands to facilitate the OpenSpec workflow. This includes checking initialization status (openspec status), viewing git history (git log), creating change containers (openspec new change), and archiving work (openspec archive). These commands are used for their intended project management purposes.
  • [PROMPT_INJECTION]: The skill implements a codebase analysis feature that scans for improvement opportunities like TODOs, missing tests, and type issues. This introduces a surface for indirect prompt injection from untrusted data within the codebase.
  • Ingestion points: The agent scans files in the codebase and git logs during Phase 2 (Task Selection) to identify potential work items.
  • Boundary markers: There are no explicit markers or instructions provided to the agent to distinguish between its own logic and potentially malicious instructions embedded in codebase comments.
  • Capability inventory: The skill is capable of modifying files in the codebase and executing shell commands via the OpenSpec CLI.
  • Sanitization: No sanitization or validation of the content discovered during codebase scanning is performed before it is used to influence the agent's task suggestions.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 4, 2026, 08:44 AM
Security Audit — agent-trust-hub — openspec-onboard