The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-provided logs, symptoms, and project briefs, creating a vulnerability surface where embedded instructions could potentially influence the agent's actions.
Ingestion points: Reads user inputs, log files, PROJECT_BRIEF.md, and application source code files (SKILL.md).
Boundary markers: No explicit delimiters are defined to isolate ingested data from the core instructions.
Capability inventory: The skill has the capability to execute shell commands, read/write local files, and perform git operations.
Sanitization: There are no explicit instructions for sanitizing or escaping the content of ingested logs or code files before processing.
[COMMAND_EXECUTION]: The skill instructs the agent to run shell commands to reproduce errors and diagnose device issues using tools like oakctl and DepthAI debug logs.
The skill explicitly mandates that privileged or persistent actions such as sudo, udev edits, or firmware updates must be confirmed by the user before execution.
[EXTERNAL_DOWNLOADS]: The agent downloads reference code from the official Luxonis GitHub repository (https://github.com/luxonis/oak-examples) to perform comparisons between user code and known-good implementations.

luxonis-troubleshoot