plaud

name

plaud

description

This skill should be used when the user explicitly mentions 'plaud', 'плод', 'плауд', 'plaud status', 'pipeline status', 'pipeline health', 'plaud classify', 'classify pending', or references _inputs/Plaud/. Provides 3 MCP tools for local pipeline operations on already-synced recordings. Do NOT trigger on generic 'recording', 'transcript', or 'meeting' without explicit Plaud mention. For raw recording access (list, transcript, audio), use the official @plaud-ai/mcp tools instead.

Plaud Pipeline

Local pipeline operations on Plaud voice recordings already synced to disk as source.md files.

Available MCP Tools (this plugin)

Tool	What it does
plaud_status	Pipeline overview — counts by stage (synced/extracted/processed); verbose mode lists per-recording table
plaud_classify	Classify recordings by category + sensitivity using configured rules (dry-run available; write-back not yet implemented)
plaud_health	Integrity checks — duplicate IDs, ghost folders (no source.md)

When NOT to use this skill — use @plaud-ai/mcp instead

For raw recording access (Plaud server-side data), use the official Plaud MCP tools:

User intent	Use official tool
"show / list / find my recordings"	list_files
"get transcript of "	get_transcript
"get AI summary / notes"	get_note
"show recording details"	get_file
"download audio of "	get_file → presigned URL
"log me into Plaud"	login

This pipeline plugin only reads local files that have already been synced — it does not call Plaud's API.

Classification Categories

Category	Description
work-meeting	Group meetings recorded on Note or matched to work calendar
work-1on1	1-1 meetings (detected by "1-1", "1on1" patterns)
personal-diary	Solo recordings on Pin — journal, reflections
personal-conversation	Multi-speaker Pin recordings with personal keywords
voice-memo	Short recordings (< 3 min solo) or unclassified
interview	Recordings with "interview" in title

Sensitivity Routing

Level	Where it goes
work	Shared / work context
personal	Personal context only
private	HOLD — user decides
unknown	HOLD — awaits classification

Pipeline Stages

1. Sync — ~/Atlas/bin/plaud-sync.py (server cron) pulls new recordings via plaud CLI → creates source.md (immutable)
2. Classify — plaud_classify assigns category + sensitivity from config rules
3. Extract — creates extract.md metadata card (interactive, user-driven)
4. Process — deep analysis with routing to appropriate destination (interactive, user-driven)

Configuration

• PLAUD_DATA_DIR environment variable (required) — path to the recordings root directory
• plaud.yaml in data dir for devices + classification rules (optional)

Plaud API auth is handled by the official @plaud-ai/cli and @plaud-ai/mcp (OAuth tokens in ~/.plaud/tokens.json). This plugin no longer reads PLAUD_TOKEN.