verification-before-completion

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No attempts to override agent behavior, bypass safety filters, or extract system prompts were detected. The instructions are purely procedural and focus on verification workflows.
  • [EXTERNAL_DOWNLOADS]: The skill does not contain any URLs, remote resource references, or instructions to download external content.
  • [REMOTE_CODE_EXECUTION]: No remote scripts, unverified dependencies, or dynamic code execution patterns are present.
  • [DATA_EXFILTRATION]: No sensitive file path access, credential harvesting, or network exfiltration patterns were identified.
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and evaluate external data (command outputs and observations) to verify completion, which is a potential surface for indirect injection.
  • Ingestion points: Command results and observations mentioned in the 'Minimal Workflow' (SKILL.md).
  • Boundary markers: Not explicitly defined in the workflow.
  • Capability inventory: No specific script-based capabilities; the skill relies on the agent's existing toolset.
  • Sanitization: Not addressed in the process guidelines.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 08:57 AM
Security Audit — agent-trust-hub — verification-before-completion