meta-assimilate
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill reads external data from project-specific configuration files in the .claude/ directory. Ingestion points: Files located at .claude/agents and .claude/commands within the project path. Boundary markers: Absent; there are no instructions to delimit or ignore potentially malicious prompts within these files. Capability inventory: The skill uses Read, Write, Edit, and MultiEdit tools to modify the agent's behavior or configuration. Sanitization: Absent; the content is assimilated without validation or escaping, making it vulnerable to indirect prompt injection.
Audit Metadata