caveman-stats
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill accesses local JSONL session logs to extract token usage metrics. While this involves reading sensitive conversation history, the behavior is intrinsic to the tool's primary function of reporting stats to the user. No evidence of network exfiltration was found.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from session logs, which constitutes a surface for processing untrusted content from previous conversation turns. 1. Ingestion points: JSONL session logs on disk. 2. Boundary markers: Absent. 3. Capability inventory: File read (session logs) and file write (lifetime-savings suffix file). 4. Sanitization: None mentioned.
Audit Metadata