github-actions-il
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONOBFUSCATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill templates process untrusted data from GitHub events, including commit messages and pull request titles. It proactively mitigates injection risks by instructing users to bind these values to environment variables instead of using direct YAML interpolation, and uses
jqto safely construct JSON payloads. - [EXTERNAL_DOWNLOADS]: The skill integrates with external APIs and CLI tools including Hebcal for Jewish calendar data, Vercel, Supabase, and axe-core. These integrations are essential for the skill's functionality and target established or domain-specific well-known services.
- [COMMAND_EXECUTION]: The skill provides numerous shell-based automation templates for GitHub Actions. These scripts include security best practices such as least-privilege permission blocks for OIDC and fail-closed logic to ensure deployment freezes are maintained if external APIs are unreachable.
- [OBFUSCATION]: The skill correctly uses the U+200F Right-to-Left Mark (RTL) character to ensure Hebrew text renders correctly in Slack and Teams notification payloads. This is a legitimate linguistic use of a control character and does not represent an attempt to hide malicious content.
Audit Metadata