360 Secure SkillHub
360 安全技能中心

Blog Embargo

核心功能:通过`publishAt`字段协调博客文章索引可见时间与Buffer社交帖子排期,避免博客上线无社交推广或社交先于博客索引的问题。典型场景:同步发布博客与社交、博客先上线后推社交、紧急无社交发布。输入为指定ISO格式时间,输出为带`publishAt`的MDX文件、排期完成的Buffer帖子,需按步骤验证发布状态。

nissan
nissan
热度 35.25
最近更新 2026-05-22 01:00:36
分类:营销依赖:代码执行 / 读写文件 / 第三方 API应用场景:博客写作 / 社交媒体 / 日程安排 / 工作流

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add blog-embargo
Metadata
name
blog-embargo
version
1.0.1
description
Coordinate publishAt embargo timing across blog index visibility and social post scheduling. Ensures blog posts don't appear in the index before social posts are queued, and social posts don't fire before the post is indexed.
author
loki
tags
["content","publishing","buffer","social","embargo"]
created
"2026-04-03T00:00:00.000Z"
lesson-source
ollama-embeddings + portkey-patterns PRs merged without publishAt, posts live before social was scheduled
metadata
{"openclaw":{"emoji":"🚦"}}

Blog Embargo Skill

Why This Exists

The blog uses a publishAt frontmatter field to support soft embargoes: post deploys to Vercel immediately on merge but is hidden from the blog index (/blog listing) until publishAt passes. The index revalidates hourly.

Without this field, posts are indexed immediately on deploy. If social posts aren't already in Buffer before the merge, you get a window where the post is live but no social amplification is queued.

This skill coordinates the three moving parts:

  1. publishAt in MDX frontmatter (blog index gate)
  2. --publish-at in convert-to-mdx.py (how to set it)
  3. Buffer scheduling at publishAt + 15 min (social timing)

The Rule

Blog index visibility and social posts must be coordinated. Always decide publishAt before raising the PR.

Scenario publishAt Buffer timing
Want coordinated launch (blog + social together) Set to future slot Schedule Buffer at publishAt + 15 min
Post can go live now, social later same day Set to now + 30 min Schedule Buffer at publishAt + 15 min
No embargo needed (emergency fix, no social) Omit or set to now Post immediately

Step-by-Step

1. Decide the publish time

Before raising the PR, decide a specific AEST datetime:

  • Check Buffer queue for conflicts (use Buffer dashboard or GraphQL query)
  • Align with content calendar (LinkedIn 10am AEST, X 8-9am AEST preferred slots)
  • Minimum: now + 30 min (enough time to schedule Buffer before deploy)
publishAt = "2026-04-04T09:00:00+11:00"   # AEDT
publishAt = "2026-04-04T09:00:00+10:00"   # AEST

2. Set publishAt in MDX

Pass --publish-at to convert-to-mdx.py:

python3 scripts/blog/convert-to-mdx.py \
  --input projects/blog-pipeline/converted/<slug>.md \
  --output projects/blog-pipeline/converted/<slug>.mdx \
  --slug <slug> \
  --title "<title>" \
  --description "<description>" \
  --tags "<tags>" \
  --author "Nissan Dookeran" \
  --publish-at "2026-04-04T09:00:00+11:00"

Check: MDX frontmatter should contain publishAt: "2026-04-04T09:00:00+11:00".

3. Raise PR and merge

Post deploys immediately. Direct URL (/blog/slug) resolves. Blog index (/blog) hides the post until publishAt.

4. Schedule Buffer at publishAt + 15 min

After merge, schedule social posts with dueAt = publishAt + 15 min:

# X/Twitter — at publishAt + 15 min
node scripts/buffer-post.mjs \
  --text "$(cat projects/social-growth/thread-<slug>.md | ...)" \
  --channel twitter \
  --publish-at "2026-04-04T09:15:00+11:00"

# LinkedIn — same time or later slot
node scripts/buffer-post.mjs \
  --text "$(cat projects/social-growth/linkedin-<slug>.md | ...)" \
  --channel linkedin \
  --publish-at "2026-04-04T09:15:00+11:00"

Or use the batch file approach with scripts/buffer-post.mjs --file posts.json.

5. Verify

  • Direct URL resolves immediately after merge ✅
  • Blog index shows post only after publishAt ✅
  • Buffer shows posts scheduled at publishAt + 15 min ✅

Recovery: Post Already Live Without publishAt

If a post was merged without publishAt and is already indexed:

  1. The embargo window is gone — post is already visible
  2. Schedule social posts for the next appropriate slot (same day if possible)
  3. Don't backdate social posts to before the post went live
  4. Note the gap in the daily memory log

This happened with:

  • ollama-embeddings (2026-04-03) — live, no social queued
  • portkey-patterns (2026-04-03) — live, no social queued

How publishAt Works (Technical)

  • publishAt is an ISO8601 field added to MDX frontmatter
  • The blog index page (/blog) filters posts where publishAt > now
  • The export const revalidate = 3600 means the index updates hourly
  • Direct slug URL (/blog/slug) always resolves — no gate at the page level
  • convert-to-mdx.py sets both publishAt and aligns the date field when --publish-at is provided

Integration Points

  • Playbook: playbooks/blog-publish/PLAYBOOK.md — Step 0b (Embargo Gate) uses this skill
  • Script: scripts/blog/convert-to-mdx.py — --publish-at flag
  • Script: scripts/buffer-post.mjs — --publish-at flag or publishAt in batch JSON
  • Buffer channels: X 69c29939af47dacb694d3d1f, LinkedIn 69c29382af47dacb694d24b4

Lessons Learned

2026-04-03: Two posts (ollama-embeddings, portkey-patterns) merged without publishAt. Both immediately indexed. Social copy existed but hadn't been approved yet, so nothing was in Buffer. Result: posts live with no social amplification window. Fix: this skill + mandatory Step 0b in blog-publish playbook.