skills/cerredz/vidbyte-skills/para/Gen Agent Trust Hub

para

Pass

Audited by Gen Agent Trust Hub on Aug 25, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill processes user-provided notes and files, representing an indirect prompt injection surface. This is mitigated through specific behavioral guidelines.
  • Ingestion points: Note identifiers and contents are ingested during Phase 5 (Classification) and Phase 6 (Review) as described in SKILL.md.
  • Boundary markers: The skill contains explicit directives to treat note contents as untrusted and private, and it identifies embedded note instructions that attempt to redirect the workflow as a violation of its use case.
  • Capability inventory: The system is authorized to read local files and write a state file (para-state.json). File movement is strictly limited to operations explicitly authorized by the user.
  • Sanitization: The interaction contract requires the agent to halt for user input at every step and forbids the use of invented analytics or metadata.
  • [COMMAND_EXECUTION]: The skill manages file system organization using agent-provided tools. The risk of unauthorized file system changes is mitigated by a mandatory confirmation gate that requires a user decision and a valid actionability justification for each file operation.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 25, 2026, 09:14 PM
Security Audit — agent-trust-hub — para