dasheng-publish-operations-bridge

name	dasheng-publish-operations-bridge
description	Use when a Dasheng Publish channel pack needs account-launch, cold-start, matrix-role, cadence, hook, keyword, series, CTA, or post-publish metric advice before platform execution. Normalizes the external agent-skills-launch-pack into a small machine-readable Publish advisory without uploading or publishing.

Dasheng Publish Operations Bridge

Role

This is the account-operations advisory layer inside publish.

It consumes account_operations_request.json, consults the matching external agent-skills-launch-pack skill, and writes normalized advice beside the channel pack. It never logs in, uploads media, edits cookies, schedules posts, or clicks a final publish button.

The publish executor remains responsible for platform delivery. This bridge only improves the decision around positioning, hook, keywords, series, cadence, interaction, experiment design, and post-publish review.

External Upstream

• Repository: https://github.com/chenjin-cmd/agent-skills-launch-pack_
• Root env: AGENT_SKILLS_LAUNCH_PACK_ROOT
• Default root: /Volumes/PSSD/agent-skills-launch-pack
• Registry: configs/publish/upstream_repos.json

Do not vendor the upstream skills into this repository or install them into global Codex skills. Read them from the external checkout so upstream updates remain easy to inspect and sync.

Platform mapping:

Publish channel	Upstream skill
wechat_article	wechat-account-launch-expert
xiaohongshu_video	xiaohongshu-account-launch-expert
douyin_video	douyin-account-launch-expert
x_post	x-twitter-cold-start-expert
future wechat_channels_video	channels-account-launch-expert

Input

Required:

• account_operations_request.json
• The referenced channel_pack.json
• The referenced upstream SKILL.md

Useful account context:

• account_stage: new, cold_start, active, low_performance, dormant, risk_review, or matrix_experiment
• account_goal
• account_slot
• matrix_role
• target_audience
• conversion_goal
• weekly_capacity
• prior post metrics or review notes

Missing context must not cause a long interview. Produce a conservative hypothesis-based advisory and list only the missing fields that materially change the recommendation.

Workflow

1. Validate that the request and channel pack belong to the same topic and channel.
2. Resolve the external root from AGENT_SKILLS_LAUNCH_PACK_ROOT, falling back to /Volumes/PSSD/agent-skills-launch-pack.
3. Read only the mapped upstream SKILL.md. Read its references/ playbook only when the request asks for a full launch plan, account audit, content calendar, or matrix experiment.
4. Treat upstream platform claims as advisory. Current platform rules, API behavior, AI labels, advertising, finance/investment wording, and account penalties require separate official-rule verification.
5. Keep the approved article/video facts and core thesis frozen. Recommend metadata, packaging, cadence, series placement, interaction, and experiment changes; do not silently rewrite or regenerate production assets.
6. Write both normalized JSON and a short human review page in the channel pack directory.

Output Contract

Write:

• account_operations_advice.json
• account_operations_advice.md

Minimum JSON shape:

{
  "schema_version": "dasheng.publish.operations_advice.v1",
  "status": "completed",
  "topic_id": "topic-id",
  "channel": "wechat_article",
  "platform": "wechat",
  "account_stage": "cold_start",
  "upstream_skill": "wechat-account-launch-expert",
  "recommendations": {
    "positioning_check": [],
    "title_or_hook": [],
    "keywords_or_tags": [],
    "series_or_collection": [],
    "cadence": [],
    "interaction_or_cta": [],
    "risk_checks": []
  },
  "experiment": {
    "hypothesis": "",
    "success_signals": [],
    "stop_or_adjust_conditions": []
  },
  "post_publish_metrics": [],
  "missing_context": [],
  "source": {
    "repo": "https://github.com/chenjin-cmd/agent-skills-launch-pack_",
    "skill_file": "",
    "local_head": ""
  }
}

recommendations must be a JSON object. status, channel, platform, and upstream_skill must match the request. A malformed or mismatched advice file does not satisfy a required operations review.

Publish Gate

• new, cold_start, low_performance, dormant, risk_review, and matrix_experiment default to required_before_execution.
• active and unspecified default to non-blocking advisory mode.
• The publish pack may still be built while review is pending, but guarded execution must remain blocked until a valid advice JSON exists and the pack is rebuilt.

Safety

• Never promise followers, views, revenue, or viral results.
• Never recommend fake engagement, bulk account abuse, hidden contact details, review bypass, plagiarism, or copyright evasion.
• For finance and investment content, add a platform-expression risk check and preserve the article/video disclaimer.
• Keep all runtime request/advice files under ~/Desktop/自媒体创作/07_发布执行/<run_id>/...; never write them into skills/ or the project root.

Upstream Check

Before changing this bridge against a newer upstream version:

python3 scripts/check_publish_upstreams.py \
  --name agent-skills-launch-pack \
  --remote

Review the upstream diff and rerun its validator/tests before changing the normalized output contract.