The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command`` syntax in SKILL.md to automatically check for the presence of Node.js, Python, and the Deepgram SDK when the skill is first loaded. These are benign environment checks designed to verify prerequisites.
[DATA_EXPOSURE]: The implementation script attempts to collect diagnostic data from system logs which may contain sensitive information. Specifically, the command grep -i deepgram /var/log/app/*.log in references/implementation.md extracts application logs. This behavior is aligned with the skill's purpose of creating a debug bundle.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (audio files and logs) to generate a bundle for external sharing, creating a potential surface for indirect injection if malicious instructions are embedded in the processed data.
Ingestion points: User-provided audio files and application logs (/var/log/app/*.log).
Boundary markers: The instructions explicitly advise the user to review the bundle for sensitive information (keys, audio bytes, transcript text) before sharing.
Capability inventory: Uses curl, tar, grep, sed, find, and the Deepgram SDK for data collection and processing.
Sanitization: Includes automated redaction of the DEEPGRAM_API_KEY environment variable from all collected files using sed regex replacement.

deepgram-debug-bundle