mnemonics
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from the repository's
.mnemonics/directory, which can act as a vector for indirect instructions. - Ingestion points: The
@historiansubagent (viamemory_recall) retrieves content from Markdown files located in the.mnemonics/{type}/{title}.mdpath, as described inSKILL.md. - Boundary markers: The skill instructions do not define explicit delimiters or instructions to treat the recalled memory content as untrusted data, increasing the risk that the agent may follow instructions embedded within a memory file.
- Capability inventory: The skill leverages the
@historiansubagent which has capabilities for file-write (memory_remember), file-read/search (memory_recall), and file-deletion (memory_forget) on the local filesystem. - Sanitization: No sanitization, validation, or escaping mechanisms are mentioned for the content being stored in or retrieved from the memory files.
Audit Metadata