The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The advanced-vercel-debug.sh script in references/evidence-collection-framework.md executes kubectl get secret vercel-secrets -o yaml. This command captures the entire secret object, including private keys and tokens in Base64. While the output is saved to a file named secrets-redacted.yaml, the command itself performs no redaction, resulting in plain-text credential exposure.
[DATA_EXFILTRATION]: The skill is designed to create a compressed 'Evidence Bundle' containing sensitive logs, network captures, and cluster secrets for 'Support Escalation'. This workflow facilitates the movement of unencrypted, highly sensitive technical data from the local environment to external third parties without sanitization.
[COMMAND_EXECUTION]: The skill requests and utilizes broad system-level permissions via allowed-tools, including tcpdump for network sniffing on port 443 and kubectl for control plane interactions. The included bash scripts automate the collection of large-scale system state data, including journalctl logs and Jaeger traces, which may contain sensitive runtime data.

vercel-advanced-troubleshooting