skills/mindfold-ai/trellis/break-loop/Gen Agent Trust Hub

break-loop

Pass

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill defines a methodical framework for internal bug analysis and documentation. It does not involve network access, sensitive credential exposure, or the execution of untrusted external code.\n- [COMMAND_EXECUTION]: The skill directs the agent to update project-specific documentation files (e.g., in .trellis/spec/) and commit those changes using version control. These operations are restricted to the local project environment and represent standard developer assistant behavior.\n- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it incorporates debugging context and potentially user-supplied bug descriptions into persistent project documentation.\n
  • Ingestion points: Context from previous debugging sessions and human-provided bug reports.\n
  • Boundary markers: Absent; the instructions do not specify delimiters to isolate external bug data from the documentation update logic.\n
  • Capability inventory: Ability to modify local documentation files in specified directories and execute Git commit commands.\n
  • Sanitization: Absent; the skill relies on the agent's internal logic to filter and summarize debugging history without explicit sanitization rules.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 21, 2026, 02:20 AM