The Agent Skills Directory

[DATA_EXFILTRATION]: The script advanced-clay-debug.sh located in references/implementation-guide.md automates the collection of sensitive system data into a single archive. This includes:
Execution of kubectl get secret clay-secrets -o yaml, which retrieves raw, base64-encoded secrets from the environment.
Full system log collection via journalctl and container logs via kubectl logs.
Capturing raw network traffic using tcpdump on port 443.
While the script labels the secret file as 'redacted', the command itself performs no redaction, creating a high risk of credential exposure if the resulting bundle is shared.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data from external data providers and Clay enrichment tables, which represents an indirect injection surface.
Ingestion points: External provider responses and Clay table row data are processed in SKILL.md (Step 2 and Step 3) to diagnose failures.
Boundary markers: The instructions lack explicit delimitation or instructions for the agent to ignore embedded commands within the data being analyzed.
Capability inventory: The skill utilizes Bash(curl:*) and Bash(node:*) along with Read and Grep, providing a functional path for exfiltration or code execution if malicious instructions are processed.
Sanitization: There is no automated sanitization of the input data; although a manual redaction step is suggested for the final output, the internal processing of raw data remains vulnerable.

clay-advanced-troubleshooting