Listen

Improve transcription accuracy over time. Learn corrections, configure STT.

Security Scan

The skill's stated purpose (learning transcription corrections) matches its content, but the runtime instructions are vague about where corrections are stored, who can access them, and whether data is sent to external services — you should clarify persistence, access, and network behavior before installing.

Purpose & Capability

Name/description align with the instructions: the skill is intended to learn and apply transcription corrections. The included config.md shows example STT provider API keys (groq/openai/local) which are reasonable for an STT-related skill, but the skill does not declare any required credentials — these provider keys appear optional/example-only.

Instruction Scope

SKILL.md tells the agent to "track transcription errors and learn corrections" and to confirm patterns after repeated occurrences, but it is silent about where corrections are stored, how long they are retained, who/what can read them, and whether learned data is transmitted to any external endpoint or synced to provider services. The wording is open-ended and grants broad discretion to the agent to collect and persist potentially sensitive transcript/correction data.

Install Mechanism

This is an instruction-only skill with no install spec and no code files — lowest-risk installation surface (nothing is written to disk by an installer).

Credentials

The skill declares no required environment variables or credentials (good). However, config.md contains example API key placeholders for STT providers (gsk_..., sk-...) — if you later configure a provider you will need to supply keys; ensure you understand where those keys are stored and who can access them.

Persistence & Privilege

always is false and there are no declared config paths. The skill states it "auto-evolves," which implies persistent state; the platform behavior for storing skill state is not described here. Autonomous invocation is allowed by default (normal), but combined with unspecified persistence this increases the need to confirm retention and access controls.

What to consider before installing

Before installing or enabling this skill, ask the publisher/platform these concrete questions: (1) Where are learned corrections stored (platform-managed store, skill file, external service)? (2) Who/what can read or export that stored data (other skills, admins, remote services)? (3) Are transcripts or corrections ever sent to third parties or STT providers automatically? If so, which endpoints and under what conditions? (4) How long is data retained and how can a user view or delete their corrections? (5) If you supply STT provider API keys, how/where are they stored (encrypted at rest?), and will the skill transmit them somewhere? If you have privacy concerns, test the skill with non-sensitive audio first, consider disabling autonomous invocation or restricting the skill, and require explicit confirmation before any network sync or provider configuration. If the publisher cannot clearly answer these, treat the skill as risky.

Active Listening

This skill auto-evolves. Track transcription errors and learn corrections.

Rules:

• Detect when transcribed text seems off (wrong context, garbled names)
• Learn from user corrections
• Confirm patterns after 2+ occurrences
• Keep entries ultra-compact
• Check config.md for OpenClaw STT setup, criteria.md for format

---

Corrections

<!-- Mistranscriptions learned. Format: wrong → right (confidence) -->

Patterns

<!-- Recurring issues. Format: "pattern" -->

Context

<!-- Domain-specific terms that get mangled. Format: "term" -->

Never

<!-- False positives to ignore -->

---

Empty sections = no corrections yet. Learn as you go.