asset-continuity-management

What it does

This skill enforces continuity as a production control system. It requires creating three living records: an asset_manifest (one row per file/unit), a source_reference_ledger (one row per source, prompt, seed, model/context, approval, or rights constraint), and a continuity_bible (canonical descriptions and locked references). It also guides recording initial project-wide metadata, asset IDs, storage locations, approvals, and escalation triggers before generation or editing.

How it works

• Establishes three records to avoid memory- or filename-based continuity:
  • asset_manifest: per asset/file
  • source_reference_ledger: per source/reference/prompt/seed/context/approval
  • continuity_bible: canonical descriptions and locked references (characters, products, locations, wardrobe, brand, voice, music, captions, motifs)
• Records minimum setup before generation: project details, asset ID pattern, storage locations, approval policy, escalation triggers.
• Recommends an ID pattern and provides examples for asset IDs and prompts.
• Defines asset_manifest fields (e.g., asset_id, status, asset_type, prompt_id, source_ids, derived_asset_ids, retention) and emphasizes not relying on filenames alone.
• Maintains a source/reference ledger with origin, custody, rights, and generation context.
• Builds a continuity bible with canonical entries for characters, products, locations, wardrobe, brand system, audio identity, avatar identity, and caption/localization style.
• Links scene/shot intent to assets and edit decisions, and includes versioning, regeneration controls, prompt/decision logs, and edit decision tracking.
• Specifies file naming and storage conventions to ensure traceability and diffs.
• Encourages metadata provenance using standards like C2PA, IPTC, PREMIS, PBCore, and WebVTT/FADGI for captions.

When to use it

Use before generating or editing media to establish a continuity control plane and record-keeping. It also applies when managing multiple assets, derivatives, localization variants, approvals, and continuity QA across complex production pipelines.

What it can touch

Not provided in the facts block beyond declaring tools: claude-code, codex, copilot, cursor. The skill expects structured JSON/YAML/TXT records, asset IDs, prompts, seeds, models, and references, and implies writing to manifests, ledgers, and continuity bible storage as part of its workflow.

Caveats

• Emphasizes explicit recording of unknown or not_provided values when sources are missing.
• Requires mapping to host pipeline schemas if they exist; otherwise, use diffable structured formats.
• Registrations, IDs, and continuity entries must be kept in project-local storage and not inferred from filenames or memory.
• No promises of outcomes; the skill specifies how to record and organize continuity data, not automatic generation results.