The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses shell command placeholders to retrieve the versions of Node.js and Python installed on the system at the time the skill is loaded.
[COMMAND_EXECUTION]: The skill generates and instructs the execution of a Bash script that collects system information (uname), tests network connectivity using curl, and reads local files.
[COMMAND_EXECUTION]: Accesses potentially sensitive local resources, including application logs (logs/*.log) and environment configuration files (.env). The skill implements a basic redaction mechanism (using sed) to hide secret values in the .env file before including them in the diagnostic bundle.
[COMMAND_EXECUTION]: Performs network requests to Clay's official API (api.clay.com) to verify API key validity and to a user-provided webhook URL to test integration connectivity.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the log files it processes.
Ingestion points: Application logs located at logs/*.log and /tmp/clay-*.log.
Boundary markers: None; the skill reads and pipes log content directly into a text file using grep and tail.
Capability inventory: The agent has access to Bash (including curl and tar), Read, and Grep tools.
Sanitization: No sanitization or filtering is performed on the log content to remove potential instructions hidden within log entries.

clay-debug-bundle