- name
SKILL.md
技能文件保留在发布者的仓库中。
在终端中粘贴复制的命令。它会把这个 SKILL.md 安装到对应的智能体中。
npx skills add notque/vexjoy-agent --skill content-engine -a cursor -y在终端中粘贴复制的命令。它会把这个 SKILL.md 安装到对应的智能体中。
npx skills add notque/vexjoy-agent --skill content-engine -a claude-code -y在终端中粘贴复制的命令。它会把这个 SKILL.md 安装到对应的智能体中。
npx skills add notque/vexjoy-agent --skill content-engine -a codex -y在终端中粘贴复制的命令。它会把这个 SKILL.md 安装到对应的智能体中。
npx skills add notque/vexjoy-agent --skill content-engine -a grok -y在 Grok Bot 聊天中发送复制的消息。Bot 会从 GitHub 获取这个 SKILL.md,并添加到你所有 Bot 共享的技能中。
Install this skill: https://github.com/notque/vexjoy-agent/tree/HEAD/skills/content/content-engine来自发布者
Repurpose anchor content into platform-native variants. This skill produces drafts only — it does not make API calls or publish content. Posting is handled downstream by x-api (single platform) or crosspost (multi-platform).
Platform-native means each variant is written from scratch for its target platform: different register (conversational on X, professional-but-human on LinkedIn, punchy on TikTok), different structure (thread vs. long-form post vs. short script vs. newsletter section), and different hook style (open fast on X, strong first line on LinkedIn, interrupt on TikTok). Shortening the same text for each platform is not adaptation — it produces content that reads identically everywhere and fails on every platform.
Establish everything needed to write platform-native variants. Do not begin writing until this phase is complete.
Required inputs:
Acquiring a transcript source asset: When the source asset is a video URL (YouTube, etc.) rather than text, extract the transcript first:
python3 scripts/video-transcript.py transcript "<video-url>" # add --timestamps or --json -o file as needed
Use the extracted transcript as the source asset for the rest of this phase.
Gate: Source asset present AND at least one target platform identified. If either is missing, ask before proceeding. Both missing means there is nothing to work with — do not guess.
Produce only the platforms the user requested. If the user says "turn this into an X thread", produce an X thread. Offer to expand to other platforms in Phase 5, but do not produce unrequested variants.
Do not write any content in this phase. Only collect and confirm inputs.
Identify the discrete, postable units inside the source asset. Each atomic idea must stand alone as a post on at least one platform without requiring the reader to know the source.
Steps:
Fewer than 3 ideas means the source is very narrow — proceed with what exists (minimum 1 is sufficient for a single platform) and note in the output file that the source yielded fewer than expected. More than 7 means the asset lacks coherence and should be split; ask the user which section to focus on.
See ${CLAUDE_SKILL_DIR}/references/phase-playbook.md for the content_ideas.md output template.
Gate: Numbered atomic ideas saved to content_ideas.md. Each is specific and standalone. The file must exist before proceeding — context is not an artifact.
Write one draft per target platform, each starting from the primary atomic idea (or specified idea) as raw material.
Every draft must be written from scratch for its platform. Do not write one version and shorten or trim it for others — audiences on each platform recognize content that was not written for them. No two platform drafts may share a verbatim sentence. If the LinkedIn draft opens with "This article covers..." or the X tweet says "New post: [title]. Key points: 1, 2, 3", that is a summary, not an adaptation. Summaries give readers no reason to stop scrolling.
Apply platform-specific rules — see ${CLAUDE_SKILL_DIR}/references/phase-playbook.md for full detail on X, LinkedIn, TikTok, YouTube, and Newsletter register/hook/structure/length/hashtag/link/CTA rules, plus the content_drafts.md output template.
Gate: One draft per target platform saved to content_drafts.md. Self-check that no two drafts share a verbatim sentence before running scripts in Phase 4. The file must exist before proceeding.
Mechanically verify drafts before delivery. Both script checks must exit 0. The gate cannot be bypassed — LLM self-assessment alone ("I reviewed the drafts and they look clean") misses hype phrases in context and cannot do verbatim comparison reliably. Run the scripts.
python3 ~/private-skills/scripts/scan-negative-framing.py content_drafts.md
See ${CLAUDE_SKILL_DIR}/references/phase-playbook.md for the full list of banned hype phrases and replacement guidance.
If exit non-zero: Identify the flagged draft(s), rewrite only the affected sections, save to content_drafts.md, re-run the check. Do not proceed to Phase 5 until exit 0.
python3 ~/private-skills/scripts/scan-negative-framing.py content_drafts.md
This check identifies any sentence appearing verbatim in two or more platform sections of content_drafts.md.
If exit non-zero: Rewrite the flagged sentence(s) in one of the two platforms where they appear. The rewrite must be platform-native — not a synonym swap. Re-run the check. Do not proceed to Phase 5 until exit 0.
Once both scripts exit 0, verify:
Gate: Both script checks exit 0. All LLM checklist items confirmed. Update content_drafts.md status from DRAFT — pending Phase 4 gate to READY. Proceed to Phase 5 only when gate passes.
Hand off clean drafts with enough context for the user or a downstream skill to act immediately. See ${CLAUDE_SKILL_DIR}/references/phase-playbook.md (Phase 5: Delivery Details) for delivery order, per-draft inclusions, downstream handoff table, optional behaviors, and artifact list.
See ${CLAUDE_SKILL_DIR}/references/phase-playbook.md for error cases: source too long, script flag unsupported, platform unspecified, ambiguous source, fewer than 3 ideas.
Repurpose source assets into platform-native social content.
---name: content-enginedescription: Repurpose source assets into platform-native social content.---# skills/content/content-engine/SKILL.md常见问题
从发布者处下载 skills/content/content-engine/SKILL.md,或复制路径后打开仓库。PlugMCP 只收录该技能,不托管文件。
发布者将它存放在 https://github.com/notque/vexjoy-agent。
Claude、Cursor、Codex 和 Grok Build 可以读取你从发布者仓库添加的 SKILL.md。在聊天中发送技能的 GitHub 文件夹链接,Grok Bot 就会安装它。