cull-release-verify
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell commands using
npm runto interact with the release management system, utilizing a version variable that may be influenced by user input or environmental context.- [EXTERNAL_DOWNLOADS]: The agent is instructed to download release artifacts, including DMGs and archives, for verification purposes. The skill mitigates risks by specifying that artifacts should be downloaded to an isolated temporary directory.- [PROMPT_INJECTION]: The skill processes untrusted external data from Git tags, Homebrew commits, and release artifacts, creating an indirect prompt injection surface. - Ingestion points: External JSON envelopes, artifact metadata, and Homebrew provenance data (SKILL.md).
- Boundary markers: The instructions call for parsing exactly one JSON envelope and using isolated temporary directories, providing some logical separation but not a robust security boundary.
- Capability inventory: Shell command execution via
npmand network access for downloading artifacts (SKILL.md). - Sanitization: No specific input sanitization or validation of the ingested external content is mentioned.
Audit Metadata