The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The advanced-clay-debug.sh script executes kubectl get secret clay-secrets -o yaml. This command retrieves Kubernetes secrets in a Base64-encoded (unredacted) format. Despite the output filename including "redacted", no actual filtering or redaction of sensitive data is performed.
[COMMAND_EXECUTION]: The skill requires the ability to execute powerful system commands including kubectl, curl, and tcpdump to gather logs, metrics, and network captures.
[DATA_EXFILTRATION]: The skill facilitates the collection of sensitive system information, including unredacted secrets and network traffic, into a single compressed archive (.tar.gz) for external review.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted logs.
Ingestion points: Pod logs and system journals are retrieved from the environment using kubectl logs and journalctl as seen in SKILL.md.
Boundary markers: No delimiters or instructions to ignore instructions within the logs are implemented.
Capability inventory: The skill possesses the capability to run system commands via kubectl, curl, and tcpdump.
Sanitization: No sanitization or validation is applied to the retrieved log data before it is handled by the agent.

clay-advanced-troubleshooting