wait-what-zh
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists of natural language instructions intended to guide the agent's output style and language. It specifically disables model invocation, which prevents the agent from calling external tools or executing code.
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and process data from project files (
CONTEXT.mdandCONTEXT-MAP.md), creating a surface for indirect instructions if those files are attacker-controlled. - Ingestion points: Data is read from
CONTEXT.mdandCONTEXT-MAP.mdas specified inSKILL.md. - Boundary markers: None; there are no explicit delimiters or warnings to ignore instructions within the context files.
- Capability inventory: No capabilities for command execution, file writing, or network access are identified;
disable-model-invocationis active inSKILL.md. - Sanitization: No sanitization or filtering of the file content is performed before processing.
Audit Metadata