using-exploration-cycle

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill uses strong directives such as 'EXTREMELY-IMPORTANT' and 'MUST route' to enforce internal workflow logic. These are used to constrain agent behavior to a specific business process rather than attempting to bypass safety filters or ignore prior instructions.
  • [DATA_EXPOSURE]: The skill references reading a local file exploration/exploration-dashboard.md and an internal SQLite database. These actions are limited to managing session state and do not involve harvesting sensitive system files or credentials.
  • [INDIRECT_PROMPT_INJECTION]: The agent ingests data from the exploration-dashboard.md file to determine workflow phases. While this presents a potential ingestion surface, the risk is mitigated by the instruction to treat the programmatic SQLite database as the 'absolute state authority' rather than the user-modifiable markdown or chat history.
  • [COMMAND_EXECUTION]: While the skill manages control flow by invoking other skills like exploration-workflow, it does not execute arbitrary shell commands or subprocesses.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 08:07 PM
Security Audit — agent-trust-hub — using-exploration-cycle