caveman
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is composed entirely of markdown instructions and YAML configuration metadata. No scripts, binaries, or automated tasks are included.
- [SAFE]: The instructions include robust safety guardrails that force the model to drop the compressed persona when providing security warnings, irreversible action confirmations, or complex multi-step instructions to prevent user misinterpretation.
- [INDIRECT_PROMPT_INJECTION]: As a stylistic transformation skill, it ingests and processes untrusted user data to produce its output. 1. Ingestion points: The skill instructions in SKILL.md are designed to process all user-supplied queries through the caveman filter. 2. Boundary markers: The skill contains explicit 'Auto-Clarity' rules to disable the persona for high-risk content. 3. Capability inventory: No file writes, network operations, or subprocess executions are present in the skill package. 4. Sanitization: None required as the skill does not execute data.
Audit Metadata