Gamma Presentations Skill
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill communicates with the official Gamma API at public-api.gamma.app and well-known logo services like Brandfetch and Logo.dev. These are legitimate endpoints required for the skill's primary functionality.
- [COMMAND_EXECUTION]: The skill references a local CLI script at .github/muscles/gamma-generator.js, which is intended for user-driven automation tasks within the workspace.
- [PROMPT_INJECTION]: The skill processes untrusted external data from local workspace files (e.g., README.md, CHANGELOG.md) and user-provided text to generate presentation content. This creates a surface for indirect prompt injection. Ingestion points include the inputText parameter and referenced workspace files. No explicit boundary markers or content sanitization mechanisms were identified in the provided templates, although the capability is limited to network writes to the Gamma API.
Audit Metadata