deepgram-java-audio-intelligence
Use when writing or reviewing Java code in this repo that enables Deepgram intelligence overlays on `/v1/listen` audio transcription - diarization, entity detection, sentiment, summarize, topics, intents, language detection, and redaction. Same endpoint as plain STT, but with extra request fields on `ListenV1RequestUrl` or `MediaTranscribeRequestOctetStream`. Use `deepgram-java-speech-to-text` for plain transcripts and `deepgram-java-text-intelligence` for analysis on existing text. Triggers include "audio intelligence", "diarize", "summarize audio", "sentiment from audio", "topic detection", and "redact".
88
82%
Does it follow best practices?
Impact
100%
1.13xAverage score across 3 eval scenarios
Advisory
Suggest reviewing before use
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 (intelligence overlays on audio transcription), lists specific capabilities, differentiates from related skills, and provides explicit trigger terms. The 'Use when' clause is present at the start, and the description includes both class-level technical details and natural language triggers. Minor note: it uses second-person-free, imperative voice which is acceptable.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: diarization, entity detection, sentiment, summarize, topics, intents, language detection, and redaction. Also names specific classes like `ListenV1RequestUrl` and `MediaTranscribeRequestOctetStream`. | 3 / 3 |
Completeness | Clearly answers both 'what' (intelligence overlays on /v1/listen audio transcription with specific features listed) and 'when' (explicit 'Use when' clause at the start, plus differentiation from related skills and explicit trigger list). | 3 / 3 |
Trigger Term Quality | Includes natural trigger terms users would say: 'audio intelligence', 'diarize', 'summarize audio', 'sentiment from audio', 'topic detection', 'redact'. Also includes domain terms like 'diarization', 'entity detection', 'language detection' throughout the description. | 3 / 3 |
Distinctiveness Conflict Risk | Explicitly distinguishes itself from `deepgram-java-speech-to-text` (plain transcripts) and `deepgram-java-text-intelligence` (text analysis), creating a clear niche for audio intelligence overlays specifically. The trigger terms are highly specific to this domain. | 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 good executable code examples and useful SDK-specific gotchas. Its main weaknesses are redundant content (analysis fields listed multiple times), lack of error handling/validation guidance, and a somewhat monolithic structure that could benefit from splitting reference material into separate files. The routing section at the top is a nice touch for disambiguation.
Suggestions
Consolidate the analysis field listings — they appear as a bullet list, then again under 'Key parameters'. Keep one authoritative list and reference it.
Add error handling guidance: show how to check response status and verify that intelligence fields (e.g., sentiment, summary) are present in the response object.
Move the extensive product doc URLs and API reference links into a separate REFERENCE.md file, keeping only the most essential link in the main skill.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is mostly efficient but includes some unnecessary content. The extensive list of product doc URLs, the repeated enumeration of analysis fields (listed once as bullet points, then again under 'Key parameters'), and the 'Central product skills' section add bulk. The 'When to use this product' routing section is helpful but slightly verbose. | 2 / 3 |
Actionability | Provides fully executable, copy-paste ready Java code for both REST and WebSocket paths, with correct imports and builder patterns. The field names are specific and verified against the repo source, and the gotchas section provides concrete, actionable warnings about SDK-specific behavior. | 3 / 3 |
Workflow Clarity | The skill clearly shows how to construct and execute requests, but there is no validation or error-handling guidance. For an API integration skill, there's no mention of checking response status, handling errors, or verifying that intelligence fields actually returned results. The workflow is essentially 'build request → call API → done' with no checkpoints. | 2 / 3 |
Progressive Disclosure | The skill references example files and external docs, but with no bundle files provided, the references to repo paths are unverifiable. The content itself is somewhat monolithic — the repeated field listings and extensive external links could be better organized into a separate reference file. However, the layered API reference section is a good structural choice. | 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-java-sdk
- Commit
de2dd4b
- 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.