vibe-delegate

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill architecture relays briefs to an external agent, creating a surface for indirect instructions. Ingestion points: The --brief file or stdin input as described in references/dispatch-and-poll.md. Boundary markers: Recommended usage of XML-like tags (e.g., , <action_safety>) in references/writing-the-brief.md to delimit instructions. Capability inventory: Execution of node shell commands, file system modifications, and network access via the vibe CLI implementer. Sanitization: Comprehensive diff review and mandatory re-running of project gates are documented in references/review-and-land.md to verify implementer output before landing.
  • [COMMAND_EXECUTION]: The skill executes a relay script using node and interacts with the vibe CLI tool to perform task delegation.
  • [EXTERNAL_DOWNLOADS]: The documentation references the installation of the mistral-vibe CLI tool, which is provided by a well-known AI services vendor.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 06:41 PM
Security Audit — agent-trust-hub — vibe-delegate