save-idea
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes
git rev-parse --show-toplevelto determine the repository root directory. This is a standard operation used to provide contextual metadata for saved ideas. - [PROMPT_INJECTION]: The skill presents an indirect prompt injection attack surface by ingesting raw user input and writing it directly to local files without sanitization or boundary markers.
- Ingestion points: User input provided via the
/save-ideacommand or related triggers described in the workflow. - Boundary markers: None; the skill appends user text directly to
VIDEO-IDEAS.mdorTOPICS.mdwithout delimiters that could signal the agent to ignore embedded instructions during future reads. - Capability inventory: The skill reads and appends to files within
~/code/content/and executes shell commands for context identification. - Sanitization: Absent; the instructions explicitly mandate that the agent must keep the user's wording verbatim, which bypasses potential sanitization steps.
Audit Metadata