AGENT LAB: SKILLS

workers-best-practices

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • EXTERNAL_DOWNLOADS (MEDIUM): The skill fetches documentation from developers.cloudflare.com and downloads the @cloudflare/workers-types package from the npm registry. While these sources are related to the skill's purpose, they are external and not on the predefined trusted list.
  • COMMAND_EXECUTION (MEDIUM): The skill instructions include the use of shell commands such as 'npm pack', 'tar', and 'npx' to retrieve and process type definitions. This represents a risk of executing or processing potentially untrusted data via the shell.
  • REMOTE_CODE_EXECUTION (MEDIUM): Execution of 'npx wrangler' and 'npx tsc' involves running code from packages that may be fetched or updated at runtime.
  • PROMPT_INJECTION (LOW): The skill ingests untrusted data from external Cloudflare documentation which could contain malicious instructions. 1. Ingestion points: developers.cloudflare.com URL mentioned in SKILL.md and references/review.md. 2. Boundary markers: Absent. 3. Capability inventory: Shell access (npm, npx, tar, mkdir). 4. Sanitization: Absent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 04:43 PM