The Agent Skills Directory

[SAFE]: The skill serves as a procedural guide for firmware verification and does not contain any malicious instructions or hidden payloads.\n- [COMMAND_EXECUTION]: The skill utilizes standard, non-malicious shell commands (e.g., idf.py, ls, grep) to verify the local development environment and check build artifacts. These commands are consistent with the skill's stated purpose of hardware testing and do not attempt to access unauthorized areas of the system.\n- [DATA_EXPOSURE_EXFILTRATION]: While the skill involves collecting logs and screenshots as evidence, it includes a robust 'Privacy' section that instructs the agent to de-identify sensitive data (MAC addresses, credentials) and seek user confirmation before sharing any artifacts, which is a defensive security best practice.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines a clear 'Evidence Contract' and strict matrix templates that serve as boundary markers for processing external hardware logs (ingestion points). It limits its capability to standard debugging tools (capability inventory) and mandates sanitization of data (de-identification instructions), effectively minimizing the surface for indirect prompt injection.

fw-hil-testing