recursive-decision-ledger
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest external data through 'loading the prior ledger' and 'capturing new information,' creating a surface for indirect prompt injection.
- Ingestion points: Untrusted data enters the context when loading JSONL or Markdown ledgers (specified in SKILL.md).
- Boundary markers: None explicitly defined for data separation within the ledger files.
- Capability inventory: The skill uses
Bash,Write,Edit, andReadtools to manage the rollout loop (specified in SKILL.md). - Sanitization: No specific sanitization or filtering of external ledger content is described.
- Mitigating Factor: The skill includes explicit 'Promotion Rules' that require user approval and staged deployments for destructive or live actions, reducing the impact of potential injections.
- [COMMAND_EXECUTION]: The skill utilizes the
Bashtool to facilitate its rollout loop and search processes. The instructions scope this usage to managing append-only ledgers and running bounded searches for decision marks.
Audit Metadata