Update to Clara

{
  "description": "Capture, transcribe, import, attribute, and review advisor voice notes, consultant debriefs, meetings, calls, and existing audio recordings with Clara Hosted Voice. Use when the user asks to start Voice Capture, transcribe an audio file, import a case-notes-audio or case-notes-voice ZIP/JSON bundle, preserve a transcript in an ordinary folder with deduplication, or add a reviewed transcript to a Clara case. Do not use to create an adaptive external-participant interview link or to revise a deck from spoken feedback.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 275
    }
  ],
  "name": "transcribe",
  "skill_md_contents": "---\nname: transcribe\ndescription: Capture, transcribe, import, attribute, and review advisor voice notes, consultant debriefs, meetings, calls, and existing audio recordings with Clara Hosted Voice. Use when the user asks to start Voice Capture, transcribe an audio file, import a case-notes-audio or case-notes-voice ZIP/JSON bundle, preserve a transcript in an ordinary folder with deduplication, or add a reviewed transcript to a Clara case. Do not use to create an adaptive external-participant interview link or to revise a deck from spoken feedback.\n---\n\n<!-- CLARA_OPENAI_ONBOARDING_BEGIN -->\nOnboarding is optional. Continue ordinary professional work immediately,\nincluding direct specialist invocation, without checking or completing a local\nonboarding profile. Missing, unfinished, inaccessible or corrupt onboarding state,\nor unavailable voice/window controls, must never block ordinary work. Do not\nautomatically start, resume or repeatedly offer onboarding.\nOnly for a user-requested tutorial or a native teaching handoff, read\n`../clara/references/local-onboarding.md`. A verified paired lesson worker\nexecutes only its bound lesson and token; never bypass tutorial validation.\nTutorial profiles, progress, examples and feedback remain local; never send a\nchange request, stamp a tutorial receipt or call hosted interviews for a tutorial.\nCurrent user requests take precedence over saved preferences.\n<!-- CLARA_OPENAI_ONBOARDING_END -->\n\n# Transcribe\n\nAfter substantive use of this workflow, read and follow the `Plugin Improvement Feedback` section in `../clara/SKILL.md`.\n\n## Output Location Rule\n\nNever write run outputs inside this Git workspace, `static/shared`,\n`protected_downloads`, or another static-site folder. Preserve bundles, audio,\ntranscripts, review files, and `codex_run_review.md` in the user's target case,\nordinary folder, or adjacent project/output folder.\n\nUse this skill for transcription-first evidence capture. The hosted service is\nthe authorized audio/transcription layer. Durable source and output files remain\nlocal after import; Codex performs speaker review and advisory interpretation\nthrough the user's existing ChatGPT plan. This is separate from `hosted-interview`,\nwhich conducts an adaptive conversation with an external participant.\n\n## Choose the Path\n\n- **Live consultant debrief:** launch Voice Capture from an existing Clara case.\n- **Existing audio:** upload the voice note, meeting, or call recording through\n  the hosted service, then import the downloaded bundle.\n- **Downloaded bundle into a Clara case:** import into `voice_sessions/`, finish\n  speaker attribution, register the reviewed transcript, and update the case\n  evidence map before downstream advisory output.\n- **Downloaded bundle into an ordinary folder:** preserve the original bundle\n  and readable transcript with deduplication; do not initialize a case merely\n  to store the transcript.\n- **Spoken feedback that should change a deck:** complete transcription and\n  attribution here, then route the reviewed evidence to `deck-correction`.\n\nIn a case using advisory lineage, the reviewed transcript creates an\n`interview_transcript` receipt with the transcript path, locator, and hash.\nEach quote or attribution used downstream becomes a claim linked to that\nreceipt. The receipt can establish that the named speaker said the quoted\nwords; it does not by itself establish that the speaker's underlying assertion\nis true. Create a separate claim and basis when Clara relies on the assertion\nas a fact. Preserve the same claim and evidence IDs when the quote moves into a\nworkpaper, memo, decision pack, Claim Basis Map, or HTML content ledger.\n\nRun dependency checks from the plugin directory before substantive work:\n\n```bash\npython scripts/check_dependencies.py\n```\n\nInstall declared requirements only when the environment permits. Do not install\npackages at runtime from within a plugin script.\n\n## Live Capture\n\nFor an initialized Clara case:\n\n```bash\npython scripts/launch_hosted_voice.py <case-dir>\npython scripts/launch_hosted_voice.py <case-dir> --browser chrome\npython scripts/launch_hosted_voice.py <case-dir> \\\n  --cookie-header-file <private-cookie.txt>\n```\n\nWith an authenticated cookie or magic link, the launcher refreshes\n`case_brief.md`, sends compact transcription context in an authenticated HTTPS\nrequest body, and opens the hosted page with an opaque short-lived token. The\ncontext is not placed in the URL. The user-bound token is an additional session\ncontrol and does not replace Mparanza authentication. Without supplied\nauthentication material, the launcher opens an explicit browser-authenticated\nfallback without reading or attaching `case_brief.md`. Use Chrome when the\nembedded browser or stale permissions block the microphone. The browser\ndownloads a local bundle when the session ends. Directly opening the hosted\nvoice URL without a plugin-created launch token and authenticated session is not\na valid run.\n\nVoice Capture is transcription-first. Sparse follow-ups may help the advisor\nfinish a debrief, but the hosted model is not the final speaker-attribution or\nadvisory authority.\n\n## Existing Audio\n\nThe hosted page accepts an existing audio recording. When browser upload is\nblocked or the file is large, use the authenticated uploader:\n\n```bash\npython scripts/upload_hosted_audio.py <case-dir> <audio-file> \\\n  --magic-link-file <private-magic-link.txt>\n\npython scripts/upload_hosted_audio.py <case-dir> <audio-file> \\\n  --cookie-header-file <private-cookie.txt>\n```\n\nAdd source metadata such as `--title`, `--interview-date`, `--participants`, and\n`--interviewer` when known. Do not place authentication secrets in chat or run\nartifacts. The uploader stores the returned bundle under the case workspace and\nnormally imports it immediately. Use `--no-import` only when the bundle must be\ninspected before registration.\n\nFor an ordinary folder that is not a Clara case, disable both case-dependent\nbehaviors and choose the transcription language explicitly when it is not\nItalian:\n\n```bash\npython scripts/upload_hosted_audio.py <target-folder> <audio-file> \\\n  --no-case-context --no-import --language ar \\\n  --magic-link-file <private-magic-link.txt>\n```\n\n`--language` is the language spoken in the recording, not the language of the\ncase or final deliverable. For an Arabic recording, use `--language ar` and\npreserve the returned Arabic transcript. A Clara case may still use English,\nItalian, French, German, or Spanish as its output language; downstream notes,\nanalysis, and deliverables follow that case language. Do not silently replace\nthe Arabic source transcript with a translation.\n\nThe uploader saves the returned bundle under\n`<target-folder>/hosted_voice_uploads/`. Import that bundle with the ordinary\nfolder importer below. If `case_manifest.json` exists, the uploader continues\nto require a valid Clara case workspace.\n\n## Import Into a Clara Case\n\nUse the newest valid download by default:\n\n```bash\npython scripts/import_latest_hosted_voice_bundle.py <case-dir>\n```\n\nPoint to a specific bundle only when necessary:\n\n```bash\npython scripts/import_hosted_voice_bundle.py <case-dir> <downloaded-bundle.zip>\n```\n\nThe importer preserves the raw payload and media, registers the transcript as\nsource material, creates a local review pack, and prevents repeated imports of\nthe same session. Treat the imported transcript as evidence, not final advice.\n\n## Import Into an Ordinary Folder\n\nWhen the target does not contain `case_manifest.json` and the user only wants a\ndurable transcript:\n\n```bash\npython scripts/import_hosted_voice_bundle_to_folder.py \\\n  <target-folder> <downloaded-bundle.zip>\n```\n\nThis path keeps or adopts the original ZIP/JSON, writes or adopts a readable\nsibling transcript, and records relative paths and SHA-256 fingerprints in\n`.clara/voice_imports.json`. Exact or repackaged duplicates reuse existing\nartifacts. Missing transcripts may be repaired. Never overwrite or delete an\nordinary-folder document; unrelated collisions receive numeric suffixes. A\nconflicting variant requires deliberate `--allow-variant` use.\n\nThis lightweight path does not create case JSON, infer speakers, register\njudgement, or promote the transcript into advisory evidence.\n\n## Speaker Attribution and Review\n\nThe hosted server transcribes audio; it is not the speaker-naming authority.\nImport may create `attributed_transcript.md` only when a single known speaker\nmakes attribution trivial. Otherwise it creates `speaker_attribution_task.md`\nand `speaker_attribution_report.json`.\n\nCodex must complete that task in the same workflow:\n\n1. Read the raw transcript, source metadata, and useful notes.\n2. Assign real names only when supported; otherwise use stable labels such as\n   `Speaker 1` and `Speaker 2`.\n3. Preserve the original unattributed transcript.\n4. Correct only obvious transcription errors whose intended wording is clear\n   from context or a trusted case glossary.\n5. Inspect for merged turns or wrong labels and keep uncertainty visible.\n\nDo not use an audio diarization model for Clara speaker attribution. Do not\nrewrite, summarize, or change meaning during the transcript-cleaning pass.\n\nFinalize a reviewed transcript with the deterministic registry helper:\n\n```bash\npython scripts/finalize_hosted_transcript.py <case-dir> \\\n  <transcript-material-id> \\\n  <voice_sessions/.../attributed_transcript.md> \\\n  --audio-pointer <source_materials/interviews/...-audio.md>\n```\n\nFinalization also creates a content-hash-stable `interview_transcript` evidence\nreceipt for the exact reviewed transcript bytes. It records that the transcript\nsupports attributed wording only; it does not promote the speaker's underlying\nassertion into a fact.\n\nWhen the reviewed transcript also requires judgement, questions, and live-issue\nupdates, Codex drafts a semantic integration plan and applies it with:\n\n```bash\npython scripts/integrate_transcript_review.py <case-dir> \\\n  --plan-json <integration-plan.json>\n```\n\nThe semantic plan includes any new `evidence_receipts`, `claims`, judgement\nentries, open-question links, and case-issue claim links. The helper applies the\nwhole plan atomically; it does not interpret the transcript. New judgement\nremains `pending` until the advisor makes the normal client-pack inclusion\ndecision, and it cannot later be approved without its canonical claim binding.\nUpdate `advisory_evidence_map.md` before the transcript changes a workpaper,\nstoryline, deck, memo, or decision pack.\n\n## Codex-Native Run UX\n\nUse a short checklist covering source selection, hosted capture or upload,\nbundle import, speaker attribution, transcript review, registry finalization,\nand case-evidence update.\n\nShow a compact Run Intake table with source recording or capture mode, target\ncase/folder, language, known speakers, source metadata, bundle path, and output\nfolder. Use a Decision Table only for unresolved material choices such as an\nambiguous target folder, a genuinely uncertain speaker boundary, or whether a\nconflicting recording is an intentional variant.\n\nDefault output policy: preserve the source bundle and audio, produce a readable\nreviewed transcript, and register it when the target is a Clara case. These are\nnot choices to propose when the user asked for the normal transcription run.\nSpeaker attribution is the approval checkpoint before advisory or deck use; do\nnot ask for ceremony when attribution is clear from inspected evidence.\n\nBefore long or write-heavy work, show an execution checkpoint naming the source,\ntarget folder, expected local artifacts, and whether hosted upload is required.\nEnd with an Artifact Card listing source bundle, audio, raw transcript, reviewed\ntranscript, attribution status, registry status, and unresolved uncertainty.\nCreate `codex_run_review.md` only for a blocked run or a repeatable import or\ntranscription gap. Never edit generated ZIPs during a run.\n"
}