docs(verify): harden verify secret handling

[codeaholicguy]

Summary

Hardens the verify skill per the skill security audit (credential handling only).

• Finding: skills.sh Snyk audit FAIL / HIGH (W007, risk 0.80) — the skill forces verbatim inclusion of command strings and outputs ("cite command, exit code, and key output"), risking secret exfiltration if output contains API keys/tokens/passwords.
• Change: new Hard Rule — never print/log/store/relay secrets; redact credential-shaped values before citing output; cite exit code + summarized output rather than raw dumps. Gate step 5 now says "key output with secrets redacted". memory store note requires redacting secrets first.
• Behavior: instruction text only; no executable code changed. Scoped to the credential finding; the npx command form is untouched.

Test plan

• Single-file diff: skills/verify/SKILL.md only
• Repo pre-push hooks (fmt:check, lint, test) passed

Generated with Devin