deepgram-js-voice-agent
Use when writing or reviewing JavaScript/TypeScript in this repo that builds an interactive voice agent via `agent.deepgram.com/v1/agent/converse`. Covers `client.agent.v1.createConnection()` / `connect()`, `sendSettings`, `sendMedia`, runtime updates, event handling, and function-call responses. Use `deepgram-js-text-to-speech` for one-way synthesis, `deepgram-js-speech-to-text` or `deepgram-js-conversational-stt` for transcription only, and `deepgram-js-management-api` for project/model admin rather than live agent runtime. Triggers include "voice agent", "agent converse", "full duplex", "barge-in", "function calling", and "agent.v1".
68
82%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is an excellent skill description that clearly defines its scope (interactive voice agent via the Deepgram converse API), lists concrete actions and API methods, provides explicit trigger terms, and proactively disambiguates from related skills. The negative guidance ('Use X for Y rather than this skill') is a particularly strong pattern for reducing conflict risk in a multi-skill environment.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: createConnection/connect, sendSettings, sendMedia, runtime updates, event handling, function-call responses, and references the specific API endpoint `agent.deepgram.com/v1/agent/converse`. | 3 / 3 |
Completeness | Clearly answers both 'what' (building interactive voice agents via the converse API with specific operations) and 'when' (explicit 'Use when' clause at the start, plus explicit trigger terms and negative guidance on when to use other skills instead). | 3 / 3 |
Trigger Term Quality | Includes natural trigger terms users would say: 'voice agent', 'agent converse', 'full duplex', 'barge-in', 'function calling', 'agent.v1'. Also differentiates from related skills by naming them explicitly, which helps with negative matching. | 3 / 3 |
Distinctiveness Conflict Risk | Exceptionally distinctive — explicitly names competing/adjacent skills (deepgram-js-text-to-speech, deepgram-js-speech-to-text, deepgram-js-conversational-stt, deepgram-js-management-api) and explains when to use those instead, making conflict very unlikely. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
64%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a solid, actionable skill with executable code examples and clear authentication/quick-start patterns. Its main weaknesses are slightly verbose framing sections, lack of explicit validation/error-recovery workflows (especially for function calling and audio streaming), and references to bundle files that don't exist. The gotchas section is genuinely useful and adds value beyond what Claude would know.
Suggestions
Add an explicit multi-step workflow for function calling: receive FunctionCallRequest → parse data.functions[] → execute → sendFunctionCallResponse → verify acknowledgment, with error handling at each step.
Trim the 'When to use this product' section to a brief bullet list without explanatory prose, and remove or shorten the 'Central product skills' section which is tangential to the agent runtime skill.
Include a reference.md or similar bundle file to back up the in-repo reference claims, or remove those references if no bundle is provided.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Generally efficient but includes some unnecessary explanation (e.g., 'When to use this product' section explains concepts Claude can infer, the 'Central product skills' section at the end is tangential, and some prose like 'Full-duplex voice agent runtime over...' restates what the description already covers). The code examples are lean, but surrounding text could be tightened. | 2 / 3 |
Actionability | Provides fully executable, copy-paste-ready code for authentication and connection setup. The quick start example is complete with real method calls, event handling, and settings configuration. Key parameters and runtime update methods are concretely named with their signatures. | 3 / 3 |
Workflow Clarity | The quick start implicitly shows a sequence (create connection → register handlers → connect → waitForOpen → sendSettings), and gotchas reinforce ordering ('Settings must be first'). However, there are no explicit validation checkpoints, no error recovery steps, and the function-call response flow is mentioned in gotchas but not shown as a complete workflow with verification. | 2 / 3 |
Progressive Disclosure | References to example files, external docs, and API specs are well-signaled and one level deep. However, there are no bundle files to support the references to 'reference.md' or the source files mentioned, and the API reference section lists 5 external links plus in-repo paths that can't be verified. The content is reasonably structured but the 'API reference (layered)' section inlines what could be a separate reference file. | 2 / 3 |
Total | 9 / 12 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 — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
- Repository
- deepgram/deepgram-js-sdk
- Commit
bcffba7
- 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.