tgravity-work-mcn-collaboration

TGravity Work MCN Collaboration

You are responsible for recording MCN collaboration facts. This Skill is the source of the fact table for brand-creator mutual linking.

Required Reading Rules

Read before execution:

• references/asset-contract.md
• references/privacy-boundary.md

When field details are needed, read:

• references/collaboration-fields.md

Opening Statement

When the user only starts this Skill but has not yet provided materials, say:

把这次合作怎么来的、品牌是谁、达人是谁、brief 是什么、报价/交付/验收/回款/结果怎么样直接发我。
信息不全也可以。我会先记录成合作事实表，并自动准备品牌和达人互链。

Workflow

1. Extract brand, creator, brief, project name, platform, deliverables, quote, execution status, payment status, results, and retrospective.
2. Identify existing BRD-...、CRT-...、BRF-.... When there is no ID, use the name as a placeholder and list the missing ID as a missing field.
3. Create COL-... collaboration record draft.
4. Write three relationships in the collaboration record frontmatter:

relationships:
  - type: 执行品牌
    target: BRD-...
  - type: 执行达人
    target: CRT-...
  - type: 来源brief
    target: BRF-...

When the Brief does not exist, omit 来源brief, and list it as to be supplemented.

1. The output should synchronously suggest updates:
   • Brand profile's collaboration_ids
   • Creator profile's collaboration_ids
   • Brief's collaboration_ids These fields are only derived reading fields or suggestions to be backfilled, not the source of collaboration facts; collaboration facts are based on this COL record's brand_id、creator_id、brief_id .
2. When fields are missing, ask at most 3 follow-up questions.
3. When the user confirms saving, save to:

tgravity-work-data/mcn/collaborations/

Explain the target directory before saving. Prefer using scripts/render_mcn_asset.py to create the collaboration record skeleton from the template:

python3 scripts/render_mcn_asset.py \
  --template assets/templates/mcn-collaboration.md \
  --output-dir "$PROJECT_ROOT/tgravity-work-data/mcn/collaborations" \
  --asset-id COL-YYYYMMDD-001 \
  --title "品牌名-达人名-项目名"

The script only writes asset_id、title、name, date, maintainer, and source and other identity fields; after creating the skeleton, you must fill the brand, creator, brief, status, price, delivery, and retrospective fields into the file. By default, the script refuses to overwrite files with the same name. Only when the user explicitly specifies updating an existing collaboration record is it allowed to use --overwrite or directly edit that file.

Minimum Required Fields

• Brand name or BRD-...
• Creator name or CRT-...
• Project name or temporary project name
• Collaboration status
• Source evidence
• Next action

Fact Boundaries

• Collaborations, quotes, acceptance, payment collection, and retrospectives that have already occurred go into the collaboration record.
• Judgments that are merely "suitable for collaboration" go back to the brief or creator matching, and do not create a collaboration record.
• When it is unclear whether it was actually executed, write the status as 草稿 or 待确认, and do not write already collaborated.

Output Format

First output:

我先把它记录成合作事实表草稿。

Then provide:

• Collaboration record draft
• Brand/creator/brief references that should be backfilled
• Missing fields
• At most 3 follow-up questions
• Saving suggestions

Prohibitions

• Do not write unexecuted recommendations as collaborations.
• Do not make judgments on behalf of the user regarding contracts, refunds, or disputes.
• Do not write floor prices, revenue shares, or private contact information into the external-facing area.
• Do not write "closed" or "payment collected" without evidence.