cypress-debugger
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes Cypress CLI commands (node_modules/.bin/cypress) and the GitHub CLI (gh). These executions are constrained by a secure shell launcher (run-artifact-reader.sh) that selects a root-owned system Python interpreter and clears the environment to prevent exploitation via malicious local shims.
- [EXTERNAL_DOWNLOADS]: The skill can download CI test artifacts from GitHub using download-cypress-reports.py. This process is limited to user-confirmed repositories and uses official GitHub API endpoints with strict path and identity validation.
- [DYNAMIC_EXECUTION]: The skill utilizes runtime code loading and system library calls. The publish-mochawesome-report.py script uses exec() to load a bundled sibling helper (read-cypress-artifact.py), while download-cypress-reports.py uses ctypes.CDLL to access atomic rename functions for secure file operations. Both are implemented with strict security checks including file ownership verification and absolute path resolution.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted test artifacts. Evidence chain: 1. Ingestion points: cypress/reports/, cypress/screenshots/, and cypress/videos/. 2. Boundary markers: SKILL.md instructions require the agent to quote strings and label output sources. 3. Capability inventory: Subprocess calls are limited to Cypress runs and GitHub CLI usage. 4. Sanitization: Redaction logic in redact_artifact.py and residual_credentials.py filters sensitive data.
Audit Metadata