media-router

Media Router

Use When

• The user has audio, video, transcript, subtitle, A-roll, or B-roll material.
• The first decision is whether to transcribe, package subtitles, add semantic understanding, or build an edit-ready plan.
• The output quality target is rough, subtitle-ready, or edit-ready.

Do Not Use When

• Do not treat every media request as plain transcription.
• Do not execute downstream collection or editing work in this router.
• Do not choose a costly model chain when rough output is enough.
• Do not route image, video, audio, or workflow generation here; use generation-router.

Core Rule

Route by job intent, not by input file type alone.

Choose from five dimensions:

• inputType: audio, video, transcript, or transcript+assets
• goal: transcript, subtitles, semantic-analysis, beats, broll-map, or edit-plan
• quality: rough, subtitle-ready, or edit-ready
• scale: single or batch
• costMode: cheap-first or quality-first

Default Routing Logic

• Audio to transcript -> audio-transcription.
• Video to transcript or subtitles -> video-transcription.
• Existing transcript to subtitle files -> subtitle-packager.
• Transcript plus assets to cut logic or B-roll plan -> editing-decision-engine.
• Video to edit-ready plan -> transcription first, optional subtitle packaging, optional semantic video analysis, then editing-decision-engine.
• Analyze first, then generate -> this router handles the understanding step, then hands back to generation-router.

Quality Thresholds

• rough: quick review, search, logging, or candidate beat spotting. Prefer cheaper transcription and fewer enrichments.
• subtitle-ready: captions, SRT/VTT, and reliable sentence or word timing. Timestamps are mandatory.
• edit-ready: A-roll/B-roll decisions, cutaway placement, and beat-level planning. Use timestamped transcript plus semantic understanding when visual proof matters.

Output Shape

The artifact contains:

• route and why
• primarySkill
• supportingSkills
• modelPlan
• outputArtifacts
• needsTimestamps
• executionMode

Stop Conditions

• Stop when required user intent, source evidence, or owned input artifacts are missing and guessing would change the result.
• If an owned CLI or script command fails, report the exact error and stop. Do not bypass the failure with metadata-only answers, readiness probing, local payload rewrites, fallback providers, or unpublished tools.

Handoff

• Hand the route JSON to the named primarySkill.
• Preserve expected artifacts so downstream skills know whether to produce transcript.json, timed-transcript.json, subtitles, beat maps, or edit plans.

Public Command Boundary

• Choose the smallest matching command or workflow from the user input and run it directly.
• If an owned CLI or script command fails, report the exact error and stop. Do not bypass the failure with metadata-only answers, readiness probing, local payload rewrites, fallback providers, or unpublished tools.
• This public skill is instruction-driven. Produce the artifact described by the workflow directly from the available evidence.
• Do not call private provider/runtime paths or unpublished local tools.
• If the CLI returns a quote-confirmation challenge, run postplus quote confirm --json --challenge-file <challenge.json> and retry with the returned token.