ClawAudit verdict
cypress-agent-skill
A production-grade Cypress E2E testing reference skill covering selectors, network stubbing, auth patterns, CI configuration, and Page Object Model; all capabilities (package_install for cypress/npx, network_in for test targets) are used legitimately for browser-based testing.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
Automated static analysis — not a human review. ClawAudit flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
Permission integrity
credential_access
package_install
Findings (4)
Possible hardcoded credential
examples/custom-commands.js · prose · downgraded · token: 'fake-token-123
Writes to SKILL.md — self-modifying skill
README.md · prose · downgraded · SKILL.md` automatically when you ask it to write
Uses exec() — may execute shell commands
references/commands.md · code · exec(
Opens WebSocket connection
references/network.md · code · WebSocket
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class B). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.
Permissions & capabilities
No declared permissions — minimal attack surface.
package_installnetwork_incredential_access Is this flag fair?
Thanks — recorded.