| name | talk-once |
| description | Turn a conversation, voice note, transcript, or guided interview into one source-grounded canonical article and reviewed channel-specific drafts. Use when someone wants to be interviewed about an idea, create long-form content from speech, repurpose one piece across channels, build a content package, or run the Talk Once workflow with an optional VoiceForge profile. |
Talk Once
Turn one substantive conversation into reusable content while keeping the user
in control of truth, voice, and publishing.
Operating rules
- Keep Markdown and JSON files as the source of truth.
- Work with any model or agent harness available to the user.
- Separate the user's claims from sourced facts and agent inference.
- Use a voice profile as style guidance only; never invent beliefs or claims.
- Establish one canonical article before producing derivatives.
- Ask for approval at the brief, article, and distribution boundaries.
- Never publish or post without explicit approval.
- Keep recordings, raw transcripts, credentials, and unpublished personal data
outside public repositories.
Workflow
- Create a content package from the Talk Once template or schema.
- If the package references a voice profile, read it before interviewing.
- Interview for the idea, audience, evidence, personal experience, useful
examples, disagreements, and desired outcome. Ask one focused question at a
time when the interface supports conversation.
- Attach supplied sources. Research only when requested or when factual
verification is necessary. Preserve links and distinguish evidence from
interpretation.
- Produce a brief containing the thesis, audience, promise, outline, claims,
sources, unknowns, and tone notes. Stop for approval.
- Draft one canonical long-form article from the approved brief. Preserve the
speaker's meaning while tightening structure and prose. Stop for approval.
- Produce only the requested derivatives from the approved canonical article.
Adapt format and length without changing the underlying position.
- Present the distribution set for approval. Do not publish automatically.
Interoperability
Prefer stable contracts over harness-specific state:
- Voice providers export a versioned JSON profile. VoiceForge supports
forge profile --output <path>.
- Talk Once packages conform to
content-package.schema.json and may reference
a voice profile by provider, path, and version.
- Harness integrations should orchestrate this workflow without changing the
contracts.
Read references/contracts.md when creating or
validating package files.
Completion
Return the created artifact paths, current package status, outstanding approval,
and any unresolved factual questions. A draft is not published content.