workflow

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill establishes an autonomous workflow for resolving bugs and executing multi-step tasks. Because it instructs the agent to 'just fix it' when given a bug report and to 'point at logs, errors' to resolve them, it creates a surface for Indirect Prompt Injection. Malicious instructions embedded in processed error logs or user-provided bug reports could potentially influence the agent's behavior during its autonomous execution phase.
  • Ingestion points: Bug reports provided by the user and system error logs processed in the 'Autonomous Bug Fixing' section.
  • Boundary markers: None identified. The instructions do not specify delimiters or safety warnings for handling external data.
  • Capability inventory: File system write access (specifically to tasks/todo.md and tasks/lessons.md) and the ability to spawn subagents for task execution.
  • Sanitization: None identified. The skill lacks instructions to sanitize or validate external inputs before they influence the planning or execution loop.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 12:16 AM