Percept Speaker ID

Identifies and tracks speakers in multi-person conversations, mapping speaker labels to names and managing voice command authorization levels.

This skill appears internally consistent, but before installing consider: (1) It assumes percept-listen and an Omi pendant—verify you actually run those components. (2) The speaker registry is a local JSON file (percept/data/speakers.json); make sure only intended, non-sensitive mappings are stored there and check file permissions. (3) Management is via a local dashboard on port 8960—ensure that dashboard is not exposed to untrusted networks. (4) Future voice-embedding functionality mentions pyannote but is not implemented—if that feature is added later it will require additional packages and potentially model/data downloads. (5) Verify the referenced GitHub repo (https://github.com/GetPercept/percept) yourself if you want source confirmation. If any of these assumptions are not true in your environment, treat the skill as incompatible or require additional review.

Type: OpenClaw Skill Name: percept-speaker-id Version: 1.0.0 The skill bundle describes a speaker identification and management system. The `SKILL.md` outlines its functionality, requirements, and how it manages speaker data in `percept/data/speakers.json`. There are no instructions for the AI agent to perform malicious actions, exfiltrate data, establish persistence, or execute arbitrary commands. The file system access mentioned is for configuration data, which is a legitimate operation for the stated purpose of the skill. No prompt injection attempts or high-risk behaviors were identified.