Plaud Recording Import

Archive selected Plaud audio, timestamped transcripts, and summaries in JishuDB with stable source associations. Keep each representation distinct and verify saved material before offering cited comparisons.

Required first step: JishuDB readiness

Start here before optional preference questions, public research, source transfer, drafting, rendering, or creating deliverables. Use the request and existing decisions for this check; apply the task-specific defaults below. If you make a task list, put JishuDB readiness first and leave dependent work pending until it passes.

1. Look for the host's JishuDB MCP tools. If available, call kb_get_capabilities({}) through the actual host connector. Reuse the selected service and existing authorization when they work.

2. If tools are absent, the connector is unconfigured, or the call fails, immediately read the bundled routine setup path. Missing tools trigger setup without requiring a failed MCP call. Reading the guide is not completion: perform its read-only machine, effective client-configuration, and supported Desktop-location checks next.

3. On a supported local machine, follow the quickstart's single discovery pass. Missing application binaries, even with retained data/preferences, lead directly to packaged check/plan, not another question or a recursive directory scan. Read the full installation contract only for a specific warning/failure; do not ask whether to generate a read-only plan. Repair an existing selected service instead of replacing it.

4. Treat plan_ready as internal progress. Reuse existing explicit installation or setup authorization when it covers the concrete transaction; do not ask again to inspect, plan, resume a bounded download, or continue after a user action. If authorization is missing, ask one short confirmation after the plan is ready. Keep the full plan, paths and hashes in tool evidence unless needed to resolve a discrepancy or requested by the user. On both platforms, run bounded download calls with the same valid plan until downloaded, then execute. download_pending means continue, not stop after reporting progress. Honor download_waiting retry intervals. Pause for actual OS, installer, agreement, OAuth or client-trust prompts; after the user acts, verify completion and continue automatically. Ask again only for uncovered material changes or genuinely missing authorization. Follow the setup guide's expiry and failure rules; never alter signed plans, bypass consent or repeatedly execute an installer with an unknown outcome.

5. Begin research, source transfer, or deliverable generation only after a real host capability call, required runtime readiness, and the task's destination and write authorization pass. A healthy endpoint, saved URL, installed app, or remembered KB is insufficient.

After installed, continue the quickstart's Connect after installation steps. Verify service health in a separate tool call after launch returns, then configure and load the connector through supported host controls yourself. Keep routine configuration work with the Agent when those controls are available; request user action only for a real presence prompt or a proven missing host capability. For supported initial-consent or owner-consent, use native OAuth without management-password setup. The user completes Trust/OAuth; allow automatic return before using a manual fallback. Do not start Connect again while the original callback is pending. A browser success page alone is not readiness. Read only the relevant onboarding or runtime-preparation section of the full setup guide when those details are needed.

Reuse setup authorization without a choice menu

Do not offer or recommend a "PDF fast path", "skip JishuDB", "local files first", or "archive later" choice, including through a question tool. JishuDB is a prerequisite of this Skill, not optional storage for its deliverables. A preference for speed or a particular output format does not waive it.

When installation approval is missing, ask whether to execute the concrete plan just produced; proceeding or postponing setup are valid choices. Do not replace that approval with an "install DB or produce local files?" menu. Do not ask about optional format, depth, variants, or style before readiness. Reuse explicit preferences from the original request; ask later only if a consequential ambiguity actually prevents the task. Necessary material-selection, source-access, transfer, and retention consent remain task-specific requirements; installation approval does not grant them.

Return USER_ACTION_REQUIRED for pending approval, OAuth, or a required client action that the Agent cannot perform through available supported controls. Return BLOCKED for unsupported or declined setup, with the observed reason and next action. If inspection or plan generation fails, report that exact failure rather than inventing a plan or offering a local-only substitute. Do not start research, transfer source material, or generate task deliverables while this prerequisite is unmet. If the user explicitly declines JishuDB and requests a different workflow, explain that this Skill is blocked; do not label a separate local-only workflow as successful execution of this Skill.

At a glance

Use this skill when

• The user wants to archive selected Plaud meetings in a JishuDB knowledge base.
• The user selects a local Plaud audio, transcript, or summary export to import.
• A resumed import needs to preserve the association between meeting representations.
• The user wants to import meetings and then compare their requirements with project documents.

Do not use this skill when

Do not use this skill to start recordings, control hardware, access unsynced device-only recordings, or import other vendors' material. If all needed material is already archived and only retrieval is requested, use jishudb-search; use jishudb for destination connection setup.

Agent workflow

Read connection and transfer before connecting or uploading, and use the provenance record for the saved association. Use the host's actual namespaced tools and live input schemas; tool names below are the unprefixed vendor and JishuDB names.

Follow the packaged task execution defaults for routine choices and bounded recovery; retain the recording-specific consent and identity rules below.

Scope and approval

1. Resolve the recording selection, material mode, destination KB, and output directory from the request. For an unqualified recording import, propose original audio plus any available existing transcript; summaries are a separate representation. Reuse an explicit user choice.
2. Reuse explicit transfer authorization when the request already identifies the selected material and destination. Otherwise disclose the concrete source-to-JishuDB transfer and obtain only the missing scope once. Do not interpret installing this Skill as account access, installation approval, or knowledge-base write approval.
3. Use the host's language for questions, reports, and meeting analysis. Ask only for ambiguous selections or necessary authorization, not for a resume, document pack, or unrelated account information.
4. Do not control hardware, start recordings, access unsynced device-only material, add other vendors, publish content, or create external tasks. Recurring imports need separately authorized host automation.

Connect and select

1. Reuse a working JishuDB MCP connection for the user's intended service. If missing or unusable, pause the task and follow the included installation and connection guide. Its platform helpers and recovery instructions are inside this package; no separate jishudb Skill or ClawHub download is required. Resolve setup paths and run helpers from references/jishudb-setup/, not the task root. Diagnose the observed failure first. Honor an already selected local or remote target; clarify an unresolved target before installing anything. Reuse existing installation approval. When this task needs writes, obtain that scope explicitly in the initial OAuth flow instead of first creating a read-only grant and immediately requesting a second authorization. Preserve OS, client-trust, agreement and device-owner prompts for the user. Resume only after the real client connection and required runtime are ready. Pending user action is USER_ACTION_REQUIRED; unsupported or declined setup is BLOCKED. Do not continue using a local-only persistence substitute.
2. Call kb_list, resolve the actual selected KB ID, and check kb_get_config and live tool schemas for document limits. Use kb_create only if the user approves creating the named destination. Do not guess a KB ID from its display name or silently select an unrelated default.
3. For cloud acquisition, call Plaud get_current_user. If authorization is missing, guide the browser login through the existing connector. Never ask for or read its token files. Confirm the intended account before reading recordings.
4. Call list_files with the relevant name/date filters, inspect candidate metadata, and use get_file for the selected ID. Resolve relative dates in the user's timezone and retain start_at separately from created_at. Inspect pagination and search coverage; an incomplete list does not prove that a meeting does not exist. If the recording has not synced, explain the vendor sync step.
5. For an approved local export, skip Plaud calls. Record that acquisition route honestly. Associate it with a vendor recording only when the user identifies that recording; otherwise use a stable user-approved local source identifier instead of inventing a Plaud ID.

Resolve prior imports before writing

1. Use the identity helper with the confirmed non-secret account scope, recording ID, and KB ID. For a local-only export, use the distinct scope local-export and its approved stable source ID. Its externalAssetId associates representations; manifestNoteId locates the durable provenance record. Do not use an email, device serial, bearer, or temporary URL as an identity input.
2. Call note_get with manifestNoteId. On NOT_FOUND only, plan a new provenance Note. Propagate authorization or service errors instead of treating them as an absent record. Before updating an existing Note, verify its Skill marker, source association, and KB match.
3. Follow kb_list_documents pagination to reconcile existing associated documents, including an interrupted import not yet recorded in the Note. Call kb_get_document for candidate IDs. Match representation and content hash; use audio.originalSha256 for original audio, not the document's potentially transcript-based sha256.
4. Reuse unchanged material. Append changed source content as an explicitly labelled revision by default; never overwrite or delete an older document automatically. If a manifest references a missing/deleted resource, report it and obtain approval before recreating it.

Acquire and import selected material

1. Before downloading audio, require mcp.upload.audio.enabled, its direct mode, the requested extension, and acceptable size/duration. Inspect speech and compressed-decoder readiness separately. Missing readiness requires repair or explicit approval to retain an original while processing remains incomplete; never promise a ready transcript in that case.
2. For audio, obtain the selected recording's current presigned_url from get_file and download actual bytes with a bounded, authorized host transfer. Keep the URL transient. Renew it through get_file after expiry; do not store it as provenance or bypass access restrictions.
3. For existing text, call get_transcript and/or get_note only for the approved representations. Preserve supplied timestamps and speaker labels. Store a summary as a summary, not a substitute transcript. Do not invoke audio recognition merely to import existing text.
4. Preserve the original downloaded/exported files in the approved output directory without overwriting existing files. Hash exact bytes. Keep collection times and import status out of the immutable source-text payload so reruns do not manufacture a new content revision.
5. Run the identity helper again with the representation and actual content SHA-256. Keep its retry identity unchanged for the same bytes and target. Execute the transfer sequence in the reference: kb_prepare_upload, its returned raw PUT, then kb_get_job. Use the shared externalAssetId for the audio and each text representation.
6. When direct upload is unavailable, supported text can use kb_upload({kbId, filename, contentBase64, externalAssetId}). Inspect its live schema first. Audio cannot use this fallback or server-side path. If the host cannot transfer audio, report the missing capability rather than silently switching to transcript-only import.
7. After uncertain completion, query kb_list_jobs using the descriptor's clientItemId before retrying bytes. Reconcile associated documents before replacing an expired preparation. A new attempt after a proven failed, uncommitted transfer must be recorded explicitly; it is not permission to duplicate an already saved recording.

Persist provenance and complete

1. Poll kb_get_job at a reasonable interval, for at most two minutes in the current interaction. A still-running job is PARTIAL, not failed or complete. Retain its ID for continuation instead of resubmitting it.
2. Call kb_get_document for each saved representation. For audio, report original retention, playback/normalization, transcription, and indexing separately. Use kb_retry_job only after inspecting the failed/cancelled job and addressing its cause. An indexing-only retry should reuse the saved transcription.
3. Read saved text with kb_read_document_text and confirm the requested source content or recognizable passages. Use kb_search in the selected KB to establish retrieval before promising cited analysis.
4. Save or merge the provenance record with note_create or note_update, using the stable Note ID and the returned updatedAt for optimistic concurrency. For creation, capture one RFC 3339 createdAt and reuse the identical request on an uncertain retry.
5. Call note_link_to_kb with the saved Note's current updatedAt. Follow its job, then note_get and kb_read_document_text for the linked document. A source-only Note or a queued link is not completed KB archival. Subsequent updates to a linked Note refresh automatically; inspect that refresh instead of linking repeatedly.
6. Only if requested, compare the imported material with the user's selected existing KB evidence. Cite returned document/chunk IDs and actual source timestamps where present. Separate original summaries from new analysis, preserve disagreements, and identify missing evidence. Never invent speaker labels or meeting commitments.

Return one state with the saved IDs, material coverage, and any required recovery action:

• COMPLETE: all requested material and provenance are saved, required processing/indexing and readback succeeded, and requested analysis is cited.
• PARTIAL: some material is saved or files are preserved, but a requested representation, processing step, or provenance refresh remains incomplete.
• USER_ACTION_REQUIRED: login, installation, trust, or another explicit user-presence gate is waiting and no partial archival needs reporting.
• BLOCKED: a required source/backend/host capability is absent or declined before archival.
• FAILED: an available operation failed before anything was archived; report the actual failure without a success-shaped substitute.

Never print credentials, upload headers, signed URLs, or account/device identifiers in the completion report. This package does not claim a tested Plaud account, WorkBuddy integration, or mainland network configuration.

Discovery

Keywords

• Chinese: Plaud录音导入、会议归档、音频导入、带时间戳转写、来源关联.
• English: Plaud import, local recording export, meeting transcript, provenance, duplicate reconciliation.

Example requests

• “把我选中的Plaud录音和已有转写导入指定知识库。”
• “导入这份Plaud本地导出，再对比已归档的项目需求。”
• “Archive selected Plaud meetings while preserving audio and transcript associations.”
• “Resume this interrupted recording import and check what is already saved.”

Nearby but different

• Read material already archived in JishuDB → jishudb-search.
• Repair the destination connection → jishudb.