castai-ci-integration
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references official GitHub Actions from trusted sources ('actions/checkout' and 'hashicorp/setup-terraform') for use in the CI/CD pipeline configuration.
- [COMMAND_EXECUTION]: The skill utilizes the 'Bash' tool to configure repository secrets using the GitHub CLI ('gh secret set').
- [INDIRECT_PROMPT_INJECTION]: The skill defines workflows that ingest data from external API endpoints ('api.cast.ai'), creating a theoretical surface for indirect injection via tool output poisoning.
- Ingestion points: Untrusted JSON data is retrieved from 'https://api.cast.ai/v1/kubernetes/clusters/' within the GitHub Action step in 'SKILL.md'.
- Boundary markers: Absent; the workflow processes the API response directly.
- Capability inventory: The skill uses the 'Bash' tool to write workflow files and manage environment secrets.
- Sanitization: The workflow uses 'jq' to extract specific numeric and status fields ('savingsPercentage', 'agentStatus'), which effectively sanitizes the input by ignoring any unexpected content or instructions in the API response.
Audit Metadata