CtrlK
BlogDocsLog inGet started
Tessl Logo

deepgram-python-conversational-stt

Use when writing or reviewing Python code in this repo that calls Deepgram Conversational STT v2 / Flux (`/v2/listen`) for turn-aware streaming transcription. Covers `client.listen.v2.connect(...)`, Flux models, end-of-turn detection. Use `deepgram-python-speech-to-text` for standard v1 ASR, `deepgram-python-voice-agent` for full-duplex interactive assistants. Triggers include "flux", "v2 listen", "conversational STT", "turn detection", "end of turn", "EOT", "listen.v2", "flux-general-en", "flux-general-multi".

91

1.82x
Quality

88%

Does it follow best practices?

Impact

95%

1.82x

Average score across 3 eval scenarios

SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

82%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A dense, well-organized reference skill: executable quick start, precise gotchas that prevent real API mistakes (string sample_rate, no language param, v1-vs-v2 close), and disciplined token efficiency. The main gaps are the placeholder mic_chunks()/elided async example and the absence of error-recovery steps for fatal errors or dropped connections.

Suggestions

Replace the mic_chunks() placeholder in the quick start with a concrete minimal source (e.g., reading linear16 frames from a file or pyaudio stream) so the example runs verbatim, and make the async snippet self-contained instead of using '...'.

Add a short error-recovery workflow: what to do on ListenV2FatalError or socket close (teardown, reconnect with backoff, whether turns resume), since long-lived streaming connections will hit these.

Move the full parameter/event enumeration into a reference file (e.g., references/params.md) and keep only the high-signal parameters inline, making the layered 'reference.md' pointer a real bundled file.

DimensionReasoningScore

Conciseness

Lean throughout — no explanation of concepts Claude already knows; a dense parameter table, terse gotchas, and code-first sections where every token earns its place. The closing 'Central product skills' section is brief navigation, not padding.

5 / 5

Actionability

The quick start is near copy-paste ready with real SDK types and correct close semantics ('send_close_stream(ListenV2CloseStream(type="CloseStream"))'), but 'mic_chunks()' is an undefined placeholder, the async block elides parameters with '...', and 'client' is only defined in a separate section.

4 / 5

Workflow Clarity

The quick start shows the full connection lifecycle (connect, register handlers, send ~80ms chunks, close, start_listening) with an error branch, but steps are not enumerated and there is no recovery guidance for 'ListenV2FatalError' or connection drops.

4 / 5

Progressive Disclosure

Clear section headers and a layered, one-level-deep API reference ('reference.md', AsyncAPI, Context7, product docs) with no nested chains. However, the parameter table and events list are fully inlined and 'reference.md' points at a repo file rather than a verifiable bundle file, leaving minor organization gaps.

4 / 5

Total

17

/

20

Passed

Description

95%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

An exemplary description: it states a precise niche, gives explicit 'Use when' guidance with a comprehensive trigger list including synonyms and model identifiers, and actively disambiguates against sibling skills. The only minor weakness is that capabilities are described as coverage topics rather than an action list.

DimensionReasoningScore

Specificity

Names several concrete capabilities — 'client.listen.v2.connect(...)', 'Flux models', 'end-of-turn detection', 'turn-aware streaming transcription' — but frames them as topical coverage ('Covers...') rather than a comprehensive list of actions, leaving minor gaps.

4 / 5

Completeness

Explicitly answers both what ('turn-aware streaming transcription' via '/v2/listen', covering connect, Flux models, EOT detection) and when ('Use when writing or reviewing Python code in this repo that calls...') with concrete trigger phrases.

5 / 5

Trigger Term Quality

The explicit trigger list ('flux', 'v2 listen', 'conversational STT', 'turn detection', 'end of turn', 'EOT', 'listen.v2', 'flux-general-en', 'flux-general-multi') covers natural phrases, abbreviations, and model identifiers comprehensively.

5 / 5

Distinctiveness Conflict Risk

Occupies a clear niche (v2 / Flux conversational STT) and explicitly routes sibling use cases to 'deepgram-python-speech-to-text' for v1 ASR and 'deepgram-python-voice-agent' for full-duplex assistants, minimizing wrong-skill triggering.

5 / 5

Total

19

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
deepgram/deepgram-python-sdk
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.