stitch-local-setup
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 a local Python script (
stitch_setup.py) to perform environment checks, launch a local setup UI, and run CLI commands. These operations are restricted to the local plugin directory and are used for administrative setup. - Evidence: Usage of
python .../scripts/stitch_setup.py check,ui, andclicommands for Windows, macOS, and Linux. - [INDIRECT_PROMPT_INJECTION]: The skill processes output from local tools (such as
list_projects) and configuration files, creating a surface for indirect prompt injection. However, the instructions include robust safeguards and explicit prohibitions against leaking sensitive data collected during these processes. - Ingestion points: The skill reads local project lists via
list_projectsand status lines fromstitch_setup.py check. - Boundary markers: Instructions explicitly tell the agent not to output API keys and to only report status lines.
- Capability inventory: The agent can execute local shell commands via the
stitch_setup.py runwrapper and access local user configuration files. - Sanitization: The skill mandates the use of hidden input prompts for credentials and warns against echoing environment variables or full environment dumps.
Audit Metadata