Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/knuckles-team/audio-transcriber/audio-transcriber-knowledge-graphnpx skills add Knuckles-Team/audio-transcriber --skill audio-transcriber-knowledge-graphgit clone --depth 1 https://github.com/Knuckles-Team/audio-transcriberWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/knuckles-team/audio-transcriber/audio-transcriber-knowledge-graph)<a href="https://agentmods.dev/skills/knuckles-team/audio-transcriber/audio-transcriber-knowledge-graph"><img src="https://agentmods.dev/badge/skills/knuckles-team/audio-transcriber/audio-transcriber-knowledge-graph.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00121 | $0.01010 |
| Opus 5 | $0.00060 | $0.00505 |
| Sonnet 5 | $0.00024 | $0.00202 |
| Haiku 4.5 | $0.00012 | $0.00101 |
Grade A, and why
audio-transcriber-knowledge-graph scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 86 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Audio Transcript → Knowledge Graph
Wire-First native ingestion: the audio-transcriber MCP server transcribes a file
and natively writes the result into the ONE epistemic-graph engine across every modality
(CONCEPT:AU-KG.ingest.enterprise-source-extractor). Ordinary transcribe_audio calls also
ingest by default; this skill covers the explicit ingest tool and the graph shape.
When to use
- Transcribe and persist an audio/video file into the knowledge graph in one step.
- Make a transcript semantically searchable (the
:Documentis chunked/embedded hub-side). - Keep the raw audio durable + deduped as a content-addressed blob tied to its transcript.
When NOT to use
- A throwaway transcription or caption export with no KG →
audio-transcriber-transcription. - Querying/searching transcripts already in the KG → the graph query/search tools
(
graph_search/graph_query), not this ingest tool.
Prerequisites & environment
Connect via the mcp-client skill against the audio-transcriber MCP server. A live
epistemic-graph engine must be reachable for anything to be written — otherwise ingestion
no-ops cleanly (the tool still returns the transcript, with ingested: null).
| Variable | Required | Notes |
|---|---|---|
WHISPER_MODEL |
optional | Default model when model is omitted |
Tools & actions
| Tool | Purpose |
|---|---|
audio_ingest_transcription |
Transcribe a file and ingest the result (blob + document + segments) |
Key parameters
audio_file— required path to the audio/video file.model— Whisper model (defaultWHISPER_MODEL/base).language— pin the language, else auto-detect.task—transcribe(default) ortranslate(English).
Graph shape
| Node / link | Meaning |
|---|---|
:MediaAsset (audio: blob) |
the raw audio bytes, content-addressed |
:Document audio:transcript:<slug> |
the full transcript text + language/duration/whisperModel |
:TranscriptSegment audio:segment:<slug>:<id> |
one timestamped span (startTime/endTime/noSpeechProb) |
:transcribedFrom |
:Transcript → source :MediaAsset |
:segmentOf |
:TranscriptSegment → its :Transcript |
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 86 lines · 121 tokens per session scan A 03aaa4479c7a
audio-transcriber-knowledge-graph is a skill published in the GitHub repository Knuckles-Team/audio-transcriber (2 stars, last pushed 8d ago), licensed MIT. It adds 121 tokens to every session and 1,010 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
jellyfin-kg-ingestion
Natively ingest a Jellyfin library into the epistemic-graph knowledge graph via the jellyfin-mcp MCP server — push library items as typed :MediaAsset/:Book nodes with :hasGenre/:performedBy/:authoredBy links and item overviews as searchable :Document nodes (jellyfiningestlibrary), and item posters as content-addressed…
archivebox-kg-ingestion
Natively ingest an ArchiveBox archive into the epistemic-graph knowledge graph via the archivebox-api MCP server — push snapshots as typed :Snapshot nodes (with :Tag + :hasTag links and per-snapshot :Document page-text) and archive results as :ArchiveResult nodes, best-effort, with the Wire-First ingest tools. Use…
whisper
OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual…
searxng-kg-ingestion
Persist SearXNG search results into the epistemic-graph knowledge graph via the searxng-mcp MCP server. Use when the agent must turn a web search into durable, semantically-searchable KG memory — each result becomes a :Document (plus :SearchQuery and :SearchEngine typed nodes with :resultOf / :fromEngine links) so…
whisper
OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual…
nextcloud-ingest
Natively ingest Nextcloud files into the epistemic-graph knowledge graph via the nextcloud-agent MCP server — fetch a file over WebDAV and store its raw bytes as a content-addressed :Blob/:AssetOccurrence plus its extracted text (pdf/office/txt or image OCR) as a linked :Document. Use when the agent must make a…