google-stitch-skills
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of external code from 'github.com/google-labs-code/stitch-skills'. The target repository is an official source for Google Labs projects.
- [COMMAND_EXECUTION]: Provides instructions for the agent to execute shell commands to verify existing installations and perform new installations via 'npx skills add'.
- [INDIRECT_PROMPT_INJECTION]: The skill processes and manages external project files like 'DESIGN.md' and 'SITE.md', which represents a potential attack surface. Ingestion points: 'DESIGN.md', 'SITE.md'. Boundary markers: None mentioned. Capability inventory: Shell command execution, file system management, and component generation. Sanitization: No explicit validation of the design files is mentioned.
- [METADATA_POISONING]: The skill claims to be an 'Official Google Stitch' project but is provided within the context of the author 'podo'. While the upstream URL is consistent with the claim, the differing author context is a metadata inconsistency.
Audit Metadata