skills/fallow-rs/fallow/sweep/Gen Agent Trust Hub

sweep

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns, unauthorized data access, or dangerous command execution were identified in the skill definition.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including user requests and pull-request states. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the data being audited. However, the instructions emphasize mapping criteria to authoritative evidence and re-running verification, which serves as a procedural guardrail.
  • Ingestion points: Processes user requests, active plans, diffs, and live pull-request states (SKILL.md).
  • Boundary markers: None explicitly defined.
  • Capability inventory: The skill directs the agent to perform verification and fix omissions, which likely involves the use of shell or git tools provided by the underlying platform.
  • Sanitization: No explicit sanitization or filtering of input data is mentioned.
  • [DATA_EXFILTRATION]: No network operations or instructions to move sensitive data to external locations were found.
  • [PROMPT_INJECTION]: The instructions do not contain attempts to override system prompts or bypass safety guidelines.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 11:12 AM
Security Audit — agent-trust-hub — sweep