meticulous-simulate-and-diff
Pass
Audited by Gen Agent Trust Hub on Mar 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use the
meticulousCLI for running simulations and standard shell commands (ls,head) for inspecting the~/.meticulous/directory. These actions are appropriate for the tool's purpose. - [PROMPT_INJECTION]: The skill processes untrusted HTML data during its analysis phase, creating a surface for indirect prompt injection. Ingestion points:
~/.meticulous/replays/**/*.metadata.json(Step 4); Boundary markers: absent; Capability inventory:meticulousCLI execution and file system access; Sanitization: absent. The risk is minimized by the skill's focus on structural diffing. - [DATA_EXFILTRATION]: Network requests are directed to the vendor's official domain (
app.meticulous.ai). The use of the$METICULOUS_API_TOKENenvironment variable follows security best practices for credential management.
Audit Metadata