muse-current
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a protocol where AI agents must treat constraints in
CURRENT.mdas a "mandatory grounding contract." This creates a surface for indirect prompt injection if malicious instructions are recorded as "hard invariants" in the shared state file. - Ingestion points: The agent reads active constraints and workstreams from
.memory/CURRENT.mdand an associated SQLite database. - Boundary markers: There are no explicit instructions to the agent to sanitize or ignore instructions embedded within the invariants it retrieves.
- Capability inventory: The skill uses
memory_currentandmemory_capturetools to append to files and interacts with a CLI tool for memory optimization. - Sanitization: No sanitization or validation of the content stored in
CURRENT.mdis specified. - [COMMAND_EXECUTION]: The skill documents CLI equivalents (
memory current,memory optimize) for manual interaction with its storage backend. These are local utility commands intended for project maintenance and data retrieval within the agent's environment.
Audit Metadata