skills/cloudflare/workerd/ci-report/Gen Agent Trust Hub

ci-report

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • Dynamic Code Execution: The skill utilizes the portal_codemode_execute tool to run custom JavaScript for gathering data from GitLab CI. While this is used to parallelize API requests and process data efficiently, executing dynamically generated code is a pattern that requires careful monitoring.
  • Command Execution: The skill provides instructions for the agent to execute GitHub CLI (gh) commands to retry failed CI runs. The skill includes guidance to confirm these actions with the user before execution, which is a recommended practice.
  • Indirect Prompt Injection Surface: The skill ingests and processes job logs from external CI systems (GitHub and GitLab) to extract error snippets. As this data is external and potentially influenced by PR contributors, it represents a surface for indirect prompt injection. The skill includes classification logic to differentiate between infrastructure flakes and material code changes to help mitigate this risk.
  • Ingestion points: GitHub Actions job logs and GitLab CI pipeline job logs.
  • Boundary markers: None explicitly defined for log processing beyond regex-based error extraction.
  • Capability inventory: Includes shell command execution (gh) and GitLab job/pipeline retry capabilities.
  • Sanitization: The skill performs basic regex filtering to extract error snippets rather than processing the entire log indiscriminately.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 06:02 PM
Security Audit — agent-trust-hub — ci-report