groove-admin-cursor-hooks
Pass
Audited by Gen Agent Trust Hub on Mar 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill writes shell scripts to the
.groove/hooks/cursor/directory and useschmod +xto make them executable. It also modifies the.cursor/hooks.jsonfile to register these scripts as automated hooks triggered by IDE lifecycle events.- [PROMPT_INJECTION]: Thecontext-reprime.shhook reads the local project file.groove/index.mdand injects its content into the AI's session context via theadditional_contextfield. This creates an indirect prompt injection surface if the local project file is modified by untrusted sources. - Ingestion points:
.groove/index.md(read during thesessionStarthook). - Boundary markers: None; the content is passed directly to the model as context without delimiters.
- Capability inventory: The skill has permission to write and execute scripts and modify IDE configurations.
- Sanitization: No sanitization or validation of the input file content is performed before injection.
Audit Metadata