mushi-health
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute various diagnostic commands using the vendor's
mushiCLI and thesupabaseCLI. These commands probe component health, list API keys, and retrieve service status. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data by querying database tables and fetching edge function logs via the Supabase MCP. This creates an attack surface where external content could theoretically influence the agent.
- Ingestion points: Supabase
query_logstool and SQL queries targeting theqa_story_runstable inSKILL.md. - Boundary markers: Absent; the skill does not define specific delimiters to isolate log content from instructions.
- Capability inventory: Local command execution using
mushiandsupabaseCLIs, and database interaction via MCP tools. - Sanitization: Not specified; the skill interprets raw log and database output for diagnostic purposes.
- [DATA_EXPOSURE]: The diagnostic process involves verifying the reachability of the
MUSHI_API_ENDPOINTand the validity of theMUSHI_API_KEYstored in~/.mushirc. It also references the use of aSUPABASE_ACCESS_TOKENfor log retrieval, which is standard for development environment health checks.
Audit Metadata