agentleFS
jonathanmalkinjonathanmalkin/jules/.claude/skills/write/SKILL.md

write

Skill65 starsChanged 7 months agoMITView on GitHub

Multi-stage article pipeline from seed to published, with optional technical depth track. Takes a topic through structured decisions, website draft, LinkedIn+X distribution, and optionally deeper technical expansion to Reddit. Two flow shapes: business-only or business+technical.

  • Commits and pushes

40 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use sec-check, read and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • waynesuttonSkill

    sec-check

    Security review checklist for Convex functions, auth logic, public queries, admin routes, webhooks, uploads, and AI-generated code. Use when reviewing code that touches user data, PII, or access control.

    Also in 3 projects that use this

  • MicrosoftSkill

    read

    Read pages from a document. Returns MinerU markdown (preferred) or PyPDF text, with optional page screenshots and sub-image fetching. When markdown is available, a figure_images_meta catalog is included — use the figures parameter to fetch specific sub-image pixels. Triggers include "read page 5", "what does page 12 say", "show me the chart on page 3", "extract pages 3-5".

    Also in 6 projects that use this

  • ninehillsSkill

    hunt

    Finds root cause before applying fixes for errors, crashes, regressions, failing tests, broken behavior, and screenshot-reported defects. Use when users report in any language errors, crashes, broken behavior, regressions, failing tests, screenshot evidence, or something that used to work and now fails. Not for code review or new features.

    Also in 4 projects that use this

  • waynesuttonSkill

    update-project-docs

    Also in 3 projects that use this

  • genomewalkerSkill

    health

    Soul system health check with remediation

    Also in 5 projects that use this

  • AnthropicSkill

    analyze

    Answer data questions -- from quick lookups to full analyses. Use when looking up a single metric, investigating what's driving a trend or drop, comparing segments over time, or preparing a formal data report for stakeholders.

    Also in 6 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

44 rules. 0 are common, 24 are distinctive, and 20 are very specific to this project.

  • 0

    Common

    Most of the 36 edited versions we read keep these.

  • 24

    Distinctive

    This author's own choices. Worth a look.

  • 20

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Check for existing drafts first:

    line 53
  2. Distinctive:

    If drafts exist, present them: "Which article?

    line 59
  3. Distinctive:

    If starting fresh, ask:

    line 61
  4. Distinctive:

    If the topic came from recent work, or [Your Name] did not supply much source material, generate a daily content packet:

    line 86
  5. Distinctive:

    Use it as supporting input, not as a substitute for real source material.

    line 92
  6. Very specific:

    If the topic looks like a repeated public question, skim Documents/Engagement/Answer-Bank/ first.

    line 101
  7. Very specific:

    Read references/decision-sprint-questions.md for the full question bank.

    line 109
  8. Very specific:

    Use AskUserQuestion for each decision.

    line 122
  9. Very specific:

    After all rounds complete, write the spec to ~/.claude/plans/article-{slug}.md:

    line 136
  10. Very specific:

    If [Your Name] supplies content for any placeholder, add it to the spec's ## Source Material section.

    line 183
  11. Very specific:

    Few-shot voice examples: Read @Profiles/Voice-Samples-Raw.md and select 1-2 samples that match the active register.

    line 216
  12. Distinctive:

    After the draft is saved, run a stop-slop structural audit on the raw AI-generated draft before presenting it to [Your Name].

    line 303
  13. Distinctive:

    If score < 35/50: flag prominently.

    line 325
  14. Very specific:

    After the structural audit, apply the scorecard in references/voice-scorecard.md.

    line 327
  15. Distinctive:

    If Truth Grounding is below 4/5, do not recommend publishing without another pass.

    line 338
  16. Distinctive:

    Wait for [Your Name] to signal completion.

    line 348
  17. Distinctive:

    After [Your Name] signals completion, read the edited file and diff against the decision spec:

    line 350
  18. Distinctive:

    Ask briefly: "You changed X, Y, Z from the spec.

    line 352
  19. Very specific:

    Read references/review-personas.md for full persona prompts.

    line 357
  20. Very specific:

    Create Code/[your-handle]/src/content/articles/{slug}.md with YAML frontmatter per references/platform-templates.md

    line 401
  21. Very specific:

    Run cd Code/[your-handle] && npm run build to verify the build passes

    line 402
  22. Distinctive:

    LinkedIn constraint: Do NOT reference Reddit or promise a technical follow-up.

    line 455
  23. Very specific:

    Write LinkedIn content to /tmp/linkedin-post.txt

    line 468
  24. Very specific:

    Write thread to /tmp/x-thread.txt (format: tweets separated by ---)

    line 475
  25. Distinctive:

    Default recommendation: Done (unless the topic is explicitly technical Story 1 content, in which case recommend continuing).

    line 484
  26. Distinctive:

    If done → skip to Stage 8.

    line 486
  27. Distinctive:

    If continuing → proceed to Stage 5.

    line 487
  28. Distinctive:

    If significant edits result, offer to re-run the panel.

    line 533
  29. Very specific:

    Update the existing article file at Code/[your-handle]/src/content/articles/{slug}.md

    line 538
  30. Distinctive:

    Update frontmatter (date, description if changed)

    line 539
  31. Very specific:

    Create new article at Code/[your-handle]/src/content/articles/{slug}-technical.md

    line 543
  32. Distinctive:

    Add cross-link in the original business article pointing to the technical version

    line 544
  33. Very specific:

    Write Reddit content to /tmp/reddit-post.md

    line 571
  34. Very specific:

    If Scripts/post-to-reddit-api.py is available and configured, use it

    line 572
  35. Distinctive:

    After confirmation, ask for the Reddit URL

    line 575
  36. Distinctive:

    Goal: Update all tracking files and move the article to published.

    line 613
  37. Very specific:

    Create/update Documents/Content-Pipeline/Drafts/{Article-Folder}/published.md with:

    line 617
  38. Very specific:

    Update Documents/Content-Pipeline/Published-URLs.md with all new URLs

    line 629
  39. Very specific:

    Update website article frontmatter platforms: field:

    line 631
  40. Distinctive:

    Add Reddit URL, LinkedIn technical URL, X technical thread URL

    line 650
  41. Distinctive:

    Add Reddit link as comment on the original LinkedIn post (remind [Your Name])

    line 651
  42. Distinctive:

    Update website article frontmatter with all platform URLs:

    line 652
  43. Distinctive:

    If the pipeline is interrupted (session ends, context clears), it can be resumed:

    line 676
  44. Very specific:

    Stage 4 decision point reached?: Check which platforms have been posted to.

    line 679

“Common” is measured: kept by at least 6 in 10 of the 36 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

@Profiles/Voice-Samples-Raw.md

Write

Single-invocation pipeline that takes a topic and produces published content across platforms, with the website as the primary canvas.

Arrives from: Direct invocation or chained from /think when content creation is identified.

Supporting references:

  • @references/content-packet-schema.md
  • @references/voice-scorecard.md
  • @references/faq-reply-loop.md
  • @references/platform-account-routing.md

Anti-Fabrication Guard

Before any platform-specific drafting: verify every factual claim in the draft is either (a) from [Your Name]'s stated source material, (b) from the article spec / creative brief, or (c) flagged with [VERIFY: ...]. Do not invent statistics, quotes, personal anecdotes, or specific claims. Use [PLACEHOLDER: personal story about X] for narrative elements [Your Name] should fill in.

Two Flow Shapes

ShapeStagesWhen
Business only1 → 2 → 3 → 4 → 8High-level concepts, outcomes, impacts. Most posts.
Business + Technical1 → 2 → 3 → 4 → 5 → 6 → 7 → 8When the topic warrants deep technical detail (code, implementation).

Stage Overview

StageWhatWhoNotes
1. Seed + Source MaterialCapture topic, stories, contextJules + [Your Name]~5 min
2. Decision Sprint + Creative BriefStructured decisions, reviewable brief[Your Name] (Jules facilitates)5-30 min
3. Write + Edit + Publish Website (Business)Draft, review panel, iterate, publishJules + [Your Name]Variable
4. LinkedIn + X (Business)Adapt, iterate, publish. Decision point.Jules + [Your Name]Variable
5. Write + Edit + Publish Website (Technical)Expand with technical depth, publishJules + [Your Name]OPTIONAL
6. RedditAdapt technical article for Reddit, publishJules + [Your Name]OPTIONAL
7. LinkedIn + X (Technical)Technical adaptations with Reddit linkJules + [Your Name]OPTIONAL
8. TrackUpdate published records, URLs, and content stateJules (auto)~1 min

Stage 1: Seed + Source Material

Goal: Understand what we're writing about and collect raw material.

Step 1: Capture the topic

Check for existing drafts first:

ls Documents/Content-Pipeline/Drafts/*/[your-handle].md 2>/dev/null

If drafts exist, present them: "Which article? Or starting fresh?"

If starting fresh, ask:

"What would you like to write about?"

From [Your Name]'s answer, Jules infers:

  1. Primary audience: Who is this for?
  2. Secondary audience: Who else might read it?
  3. Story track: 1 (Claude Code) / 2 (Build Where They Won't) / 3 (Solo Founder)

Confirm the inferred details: "Sounds like this is a Story [N] piece aimed at [audience]. That right?"

Step 2: Source material

Always ask:

"Any personal stories, experiences, or anecdotes you want in this one? Things only you could tell. I'll use placeholders for anything you don't have ready, and you can fill them in after seeing the draft."

Default message (always):

"I'll use [STORY NEEDED: ...] placeholders for personal stories. You can fill them in after seeing the draft."

Store all provided source material for the decision sprint.

Step 3: Pull the daily content packet if useful

If the topic came from recent work, or [Your Name] did not supply much source material, generate a daily content packet:

python3 .claude/scripts/generate-content-packet.py --stdout

Use it as supporting input, not as a substitute for real source material. The content packet should help identify:

  • the strongest anchor article candidate
  • recent proof points
  • likely LinkedIn and X angles
  • good prompting questions for [Your Name]

Step 4: Check the answer bank when the topic is a recurring question

If the topic looks like a repeated public question, skim Documents/Engagement/Answer-Bank/ first. Reuse any verified framing, proof pack, or canonical link direction that already exists.


Stage 2: Decision Sprint + Creative Brief

Goal: Make all the key decisions about the article upfront, before writing.

Read references/decision-sprint-questions.md for the full question bank.

Complexity assessment

Based on the topic, classify:

  • Simple (clear topic, obvious structure): 1 round, ~5-8 decisions, ~5 min
  • Medium (some structural choices): 2 rounds, ~8-14 decisions, ~10 min
  • Complex (multiple angles, audience tension): 3-4 rounds, ~14-30 decisions, ~15-30 min

Announce the classification: "This feels [simple/medium/complex]. I'll run [N] rounds of questions."

Running the sprint

Use AskUserQuestion for each decision. For opening hooks and structural choices, use previews so [Your Name] can see the options side by side.

Round 1 (always): Structure + Content questions. Opening hook, section architecture, what to expand/cut, depth level, key examples.

Adversarial challenge (always, after Round 1): "A skeptical reader sees this headline. What would they say is obvious or not worth reading?" Present the objection and a proposed counter. If the counter is weak, sharpen the angle before proceeding. See references/decision-sprint-questions.md for the full prompt.

Round 2 (medium+): Voice + Format questions. Register, visual elements, audience context, length target, code blocks.

Round 3 (complex+): Specifics. Principles/takeaways, closing format, people to tag, CTA.

Round 4 (only if needed): Gap check. Reconcile all decisions, identify conflicts, resolve ambiguity.

Save the decision spec

After all rounds complete, write the spec to ~/.claude/plans/article-{slug}.md:

# Article Spec: {Title}

Created: {YYYY-MM-DD}
Story: {track}
Topic: {topic}
Audience: {primary} / {secondary}
Complexity: {simple/medium/complex}
Length target: {short/medium/long}

## Decisions

### Structure
- Opening hook: {chosen hook with full text}
- Sections: {outline}
- Depth: {level}
- Examples: {key examples}
### Voice
- Register: {register}
- Visual elements: {list}
- Code blocks: {yes/no/count}

### Specifics
- Key takeaways: {list}
- Closing format: {format}
- Tags: {people to tag}
- CTA: {call to action}

## Source Material (if provided)
{Stories, anecdotes, data, or references [Your Name] supplied during Stage 1-2}

## Supporting Packet
{Daily content packet highlights, answer-bank pointers, or prior published links used to ground the draft}

## Placeholder Map
| Section | What's needed | Type | Priority |
|---------|--------------|------|----------|

Also save a copy to Documents/Field-Notes/Plans/{YYYY-MM-DD}-Article-{Slug}.md ([Your Name]'s review copy).

Placeholder map

After saving the decision spec, fill in the Placeholder Map section of the spec showing where personal stories, specific examples, data points, or anecdotes will be needed in each section. Present the spec (including the placeholder map) inline so [Your Name] can supply content before writing starts.

If [Your Name] supplies content for any placeholder, add it to the spec's ## Source Material section. Anything not supplied will become [STORY NEEDED: ...] placeholders in the draft.

Confirm with [Your Name]: "Decision sprint complete. Here's the spec and placeholder map: [display inline]. Want to fill any of these before I draft, or should I placeholder them all?"

Creative Brief

Synthesize all decisions into a single reviewable document. The last checkpoint before drafting.

ElementContent
WhatAsset type and format
WhoTarget audience + where they are in their journey
Core ideaOne sentence topic
Angle/hookWhat makes this worth reading
Key proof pointsEvidence, examples, data
Goal/CTAWhat the reader should do or think after
Voice notesOnly if deviating from default voice profile
ConstraintsWord count, platform requirements, things to avoid

Present the brief. [Your Name] confirms, adjusts, or redirects. Once approved, the brief becomes the input to Stage 3.


Stage 3: Write + Edit + Publish Website Article (Business Level)

Goal: Produce a business-level article for [your-domain], iterate until approved, then publish.

3a: Draft

Dispatch the content agent (Opus) with:

  1. The approved creative brief (primary input) and full decision spec (reference)
  2. Voice profile: @Profiles/Voice-Profile.md
  3. Few-shot voice examples: Read @Profiles/Voice-Samples-Raw.md and select 1-2 samples that match the active register. Use this register-to-sample mapping:
    • Technical register: Sample 1 (Reddit Change Monitor), Sample 4 (OpenClaw Tone Correction), Sample 17 (8 Layers of Memory)
    • Teaching register: Sample 2 (FetLife Apology), Sample 8 (Brand Voice), Sample 13 (Kink Education Vision)
    • Professional short-form: Sample 5 (LinkedIn Profile), Sample 15 (Content Strategy), Sample 16 (Reddit Article Direction)
    • Story 3 / Reflective: Sample 21 (Value Stack Ascent), Sample 22 (Slot Machine Brain) Pick the 2 most relevant samples for the article's topic within the register. Include [Your Name]'s version only (not the AI version). Keep each excerpt under 300 words; use the most representative passage if longer.
  4. Any source material [Your Name] provided in Stages 1-2
  5. Any useful content-packet or answer-bank material from Stage 1
  6. Instruction to the content agent:
Write a business-level article for [your-domain]. This is the primary version targeting
a business/professional audience — high-level concepts, outcomes, and impacts. A technical
deep-dive may follow as a separate piece, so don't try to cover implementation details here.

Mode: draft
Register: {register from spec}

## Your role

You are a synthesizer, not a creator. Your job is to build the structure, voice, and argument.
[Your Name] supplies the facts, stories, and personal experiences. When you don't have a real
fact or story, use a placeholder. Never invent one.

## Content classification

You CAN generate freely (no placeholder needed):
- General industry knowledge and commonly known facts
- Logical arguments, analysis, and synthesis of provided material
- Definitions, frameworks, and explanatory structure
- Transitions, introductions, and conclusions built from the spec
- Hypothetical examples clearly framed as illustrative ("Imagine...", "Consider a scenario...")

You MUST use a placeholder for:
- Personal anecdotes, stories, or experiences attributed to [Your Name]
- Attributed quotes from specific people
- Specific events, dates, or incidents not in the source material
- Statistics, metrics, or data points you cannot verify
- Names of people or organizations not provided in the spec or source material

## Placeholder format

Use these structured placeholders. The guidance fields help [Your Name] write the real content:

[STORY NEEDED: {what the story should accomplish}
  Purpose: {why this story is here, what it proves}
  Ideal shape: {length, structure, e.g. "2-3 sentences, problem-then-insight"}
  Example angle: {a direction [Your Name] could take, not the story itself}]

[FACT NEEDED: {what data or fact would strengthen this point}]
[STAT NEEDED: {what metric or number is needed, and why}]
[EXAMPLE NEEDED: {what kind of real example would work here}]

## Voice

You have been given 1-2 authentic writing samples from [Your Name] in this register. These are
your primary voice reference. Write in this style. Match the rhythm, sentence length variation,
opening patterns, and closing patterns you see in the samples.

Also apply Voice-Profile.md patterns for the {register} register. Before writing, quote 2-3
specific patterns from the samples you'll reproduce (not describe, reproduce).

## Self-review before returning

1. Does the opening match the chosen hook style?
2. Does the close match the chosen format?
3. Are there em-dashes? (Remove them.)
4. Does the depth match the spec?
5. Is the repo link footer present?
6. Would the first paragraph work as a standalone summary?
7. **Fabrication check:** Am I stating something as if [Your Name] experienced it? Is every first-person claim traceable to the source material? If not, convert to a placeholder.

## Output

Save the completed draft to: Documents/Content-Pipeline/Drafts/{Article-Folder}/[your-handle].md
Create the article folder if it doesn't exist.
Also return the full draft inline with word count and reading time.

At the end, include a placeholder summary (removed before publishing):

Placeholders remaining: {count}
- {section}: {placeholder type and brief description}
- {section}: {placeholder type and brief description}

3b: Stop-slop pre-edit audit

After the draft is saved, run a stop-slop structural audit on the raw AI-generated draft before presenting it to [Your Name]. This catches AI patterns at their source.

Dispatch a general-purpose agent (Sonnet) with:

  1. The draft from Drafts/{Article-Folder}/[your-handle].md
  2. Reference files: .claude/skills/stop-slop/references/structures.md and .claude/skills/stop-slop/references/phrases.md
  3. The decision spec (for source material cross-reference)
  4. Instruction: scan for structural AI patterns (false agency, binary contrasts, dramatic fragmentation, copula avoidance, formulaic structure, synonym cycling, rule-of-three). Score on 5 dimensions: Directness, Rhythm, Trust, Authenticity, Density (each 1-10).

Additionally, scan for untraced first-person claims: any sentence where the author says "I built/tried/found/discovered/noticed" or makes a specific factual claim (a number, a date, a named tool, a metric) that cannot be traced to the decision spec's source material section. Flag each as:

FABRICATION RISK: "{exact quote}" — no source material match found

This is not a judgment call. If the claim isn't in the source material and isn't a placeholder, flag it. Let the author decide.

Present the stop-slop findings inline alongside the draft:

## Pre-Edit Structural Audit (score: XX/50)

- [pattern]: "exact quote" → suggested fix

If score < 35/50: flag prominently.

After the structural audit, apply the scorecard in references/voice-scorecard.md. Report:

Voice scorecard: {total}/25
- Voice Match: X/5
- Truth Grounding: X/5
- Technical Specificity: X/5
- Authenticity: X/5
- Platform Fit: X/5

If Truth Grounding is below 4/5, do not recommend publishing without another pass.

3c: Author edit

Display the full draft inline, followed by the stop-slop findings.

Tell [Your Name]:

"Draft is at Documents/Content-Pipeline/Drafts/{Article-Folder}/[your-handle].md. Stop-slop findings are above. Edit directly, then let me know when you're done."

Wait for [Your Name] to signal completion.

After [Your Name] signals completion, read the edited file and diff against the decision spec:

  1. Identify intentional departures (sections removed, structure changed, tone shifted)
  2. Ask briefly: "You changed X, Y, Z from the spec. Intentional?"
  3. Store confirmed departures for the review panel.

3d: Review panel

Read references/review-personas.md for full persona prompts.

Reviewer dispatch: Based on the story track, dispatch reviewers as parallel content agents (subagent_type: "content"):

StoryReviewersModel overrides
1 (Claude Code)Voice editor + Technical peer + Anthropic PMopus, sonnet, opus
2 (Build Where They Won't)Voice editor + Outsider reader + Indie hackeropus, sonnet, opus
3 (Solo Founder)Voice editor + Outsider reader + General tech + Technical peeropus, sonnet, opus, sonnet

Each reviewer agent gets:

  • The edited draft (from 3c)
  • The decision spec (from Stage 2)
  • The list of intentional departures (from 3c)
  • Their persona prompt (from references/review-personas.md)

Stop-slop post-edit audit (parallel with reviewers): Second pass on the edited version. Dispatch a general-purpose agent (Sonnet) to scan for AI patterns that survived edits or were introduced during editing. Include delta from pre-edit score.

Synthesize reviews: Categorize findings as Critical / High-Value / Worth Considering. Present to [Your Name]:

## Review Summary

### Critical (must fix)
- [finding with reviewer attribution]

### High-Value (recommended)
- [finding with reviewer attribution]

### Worth Considering (author's call)
- [finding with reviewer attribution]

### Structural Audit (score: XX/50, delta: +/-N)
- [findings]

Fix critical issues automatically (or with confirmation if ambiguous). [Your Name] decides on the rest.

If significant edits result from the review panel: Offer to re-run the panel. [Your Name] decides.

3e: Publish to website

Once the article is approved:

  1. Create Code/[your-handle]/src/content/articles/{slug}.md with YAML frontmatter per references/platform-templates.md
  2. Run cd Code/[your-handle] && npm run build to verify the build passes
  3. Stage the specific article file: git add Code/[your-handle]/src/content/articles/{slug}.md
  4. Commit: git commit -m "article: {title}"
  5. Push: git push origin main
  6. Cloudflare auto-deploys. Capture URL: [your-domain]/articles/{slug}/

Stage 4: LinkedIn + X (Business Level)

Goal: Adapt the approved website article for LinkedIn and X, then publish.

4a: Write adaptations

Dispatch the content agent to write both platform variants. Each gets:

  • The approved [your-handle].md
  • The website URL (for linking)
  • Platform-specific format guidelines from @references/platform-templates.md
  • Platform-specific editorial guidance from @references/platform-writing-guide.md
  • LinkedIn algorithm reference from @references/linkedin-reference.md
  • Voice profile for voice consistency
  • The anti-fabrication constraint below

Anti-fabrication constraint (include in dispatch):

## Anti-fabrication constraint

You are adapting an approved article, not writing new content. Every factual claim,
personal anecdote, statistic, and named example in your adaptation must come from
the approved source article. You may:
- Reframe, condense, or restructure existing claims
- Drop claims that don't fit the platform
- Add platform-appropriate framing ("I wrote about this in depth on my site")

You must NOT:
- Introduce new personal stories not in the source article
- Add statistics, metrics, or data points not in the source article
- Attribute quotes or experiences not in the source article
- State new first-person claims ("I found...", "In my experience...")
  that aren't in the source

If you need to add context the source article doesn't provide, use:
[CONTEXT NEEDED: what would strengthen this adaptation]

LinkedIn format choice: Based on the article content, choose the best format:

  • Long text post (1,300-1,900 chars) — default for article adaptations. Technical depth welcome.
  • Short text post (100-300 chars) — teaser/conversation starter pointing to the article.
  • Document carousel (8-12 slides, PDF) — when the article has a clear step-by-step or list structure. Highest engagement format.

LinkedIn can carry technical depth — it's not limited to business-level framing. Show the system AND the result. Code snippets welcome when they serve the story.

LinkedIn constraint: Do NOT reference Reddit or promise a technical follow-up. The Reddit post doesn't exist yet. LinkedIn goes out with the website URL only.

Save variants to Documents/Content-Pipeline/Drafts/{Article-Folder}/:

  • LinkedIn.md
  • X-thread.md

4b: Review and iterate

Present both adaptations to [Your Name]. Iterate if needed.

4c: Publish

LinkedIn:

  1. Write LinkedIn content to /tmp/linkedin-post.txt
  2. Dry-run: python3 Scripts/linkedin-post.py --file /tmp/linkedin-post.txt --dry-run
  3. Show preview. Post after [Your Name]'s approval: python3 Scripts/linkedin-post.py --file /tmp/linkedin-post.txt
  4. Capture the LinkedIn URL
  5. Tell [Your Name]: "Add the article URL as the first comment on LinkedIn (LinkedIn penalizes links in body text)."

X Thread:

  1. Write thread to /tmp/x-thread.txt (format: tweets separated by ---)
  2. Show thread preview. Confirm with [Your Name]: "Thread looks good?"
  3. Post first tweet: xurl post "first tweet text", then chain replies: xurl reply TWEET_ID "next tweet" for each subsequent tweet
  4. Capture the thread URL from output

4d: Decision point

"Business-level article is live on the website, LinkedIn, and X. Want to add technical depth (expanded website article + Reddit)? Or done here?"

Default recommendation: Done (unless the topic is explicitly technical Story 1 content, in which case recommend continuing).

If done → skip to Stage 8. If continuing → proceed to Stage 5.


Stage 5: Write + Edit + Publish Technical Website Article (OPTIONAL)

Goal: Expand the article with full technical depth, iterate, and publish to the website.

5a: Scope the technical expansion

Before drafting, ask:

"Expand the existing article or publish as a separate technical deep-dive?"

Record the answer. This affects Stage 8 tracking:

  • Overwrite: same URL, updated frontmatter
  • Separate: new slug, new URL, tracking records two website URLs

5b: Draft

Dispatch the content agent (Opus) with the same framework as Stage 3 (including few-shot voice examples from Voice-Samples-Raw.md, prioritizing Technical register samples), but with updated instruction:

Expand this business-level article into a full technical deep-dive. Add:
- Code examples (sanitized — no credentials, no personal data)
- Implementation details, architecture decisions, trade-offs
- As much depth as needed — no artificial length limits
- Keep concepts concise, but don't hold back on detail

The audience is technical practitioners who want to understand HOW, not just WHAT.

Provide the approved business article as the starting point.

5c: Stop-slop pre-edit audit

Same pattern as Stage 3b. Dispatch general-purpose agent (Sonnet) to scan the raw draft. Includes the source-traceability scan for untraced first-person claims (same instruction as 3b).

5d: Author edit

Same pattern as Stage 3c. Present draft + stop-slop findings. Wait for edits. Spec reconciliation.

5e: Review panel

Same pattern as Stage 3d. Full reviewer dispatch + stop-slop post-edit audit.

If significant edits result, offer to re-run the panel.

5f: Publish to website

If overwriting:

  1. Update the existing article file at Code/[your-handle]/src/content/articles/{slug}.md
  2. Update frontmatter (date, description if changed)
  3. Build, commit, push (same sequence as Stage 3e)

If separate page:

  1. Create new article at Code/[your-handle]/src/content/articles/{slug}-technical.md
  2. Add cross-link in the original business article pointing to the technical version
  3. Build, commit, push
  4. Capture new URL: [your-domain]/articles/{slug}-technical/

Stage 6: Reddit (OPTIONAL)

Goal: Adapt the technical website article for Reddit and publish.

6a: Write Reddit adaptation

Dispatch the content agent with:

  • The approved technical article
  • The website URL(s)
  • Reddit format guidelines from @references/platform-templates.md
  • Reddit editorial guidance from @references/platform-writing-guide.md
  • The anti-fabrication constraint from Stage 4a (same block, included in dispatch)

Save to Documents/Content-Pipeline/Drafts/{Article-Folder}/Reddit.md

6b: Review and iterate

Present to [Your Name]. Iterate if needed.

6c: Publish

  1. Write Reddit content to /tmp/reddit-post.md
  2. If Scripts/post-to-reddit-api.py is available and configured, use it
  3. Otherwise: copy to clipboard with printf '%s' "$(cat /tmp/reddit-post.md)" | pbcopy
  4. Tell [Your Name]: "Reddit post copied to clipboard. Paste into r/ClaudeCode."
  5. After confirmation, ask for the Reddit URL

Stage 7: LinkedIn + X Technical (OPTIONAL)

Goal: Post technical-level content to LinkedIn and X with link back to Reddit.

7a: Write adaptations

Dispatch the content agent to write (include the anti-fabrication constraint from Stage 4a in each dispatch):

  • LinkedIn technical adaptation (references the Reddit post and/or technical website article)
  • X thread with technical framing + link to Reddit

Save to Documents/Content-Pipeline/Drafts/{Article-Folder}/:

  • LinkedIn-technical.md
  • X-thread-technical.md

7b: Review and iterate

Present both to [Your Name]. Iterate if needed.

7c: Publish

LinkedIn:

  1. Write to /tmp/linkedin-technical-post.txt
  2. Dry-run, show preview, post after approval
  3. Capture URL

X Thread:

  1. Write to /tmp/x-thread-technical.txt
  2. Dry-run, confirm, post
  3. Capture URL

Stage 8: Track

Goal: Update all tracking files and move the article to published.

Business-only flow (Stages 1-4)

  1. Create/update Documents/Content-Pipeline/Drafts/{Article-Folder}/published.md with:
    • Website URL
    • LinkedIn URL
    • X thread URL
    • Flow shape: Business only
    • Dates for each platform
  2. Move the article folder:
    mv "Documents/Content-Pipeline/Drafts/{Article-Folder}" "Documents/Content-Pipeline/Published/{Article-Folder}"
  3. Update Documents/Content-Pipeline/Published-URLs.md with all new URLs
  4. Update website article frontmatter platforms: field:
    platforms:
      linkedin: "{url}"
      x_thread: "{url}"
  5. Stage and commit:
    git add "Documents/Content-Pipeline/Published/{Article-Folder}/"
    git add Documents/Content-Pipeline/Published-URLs.md
    git add "Code/[your-handle]/src/content/articles/{slug}.md"
    git commit -m "track: {title} published (business)"

Business + Technical flow (Stages 1-7)

Same as above, plus:

  • Record second website URL if a separate technical page was created
  • Add Reddit URL, LinkedIn technical URL, X technical thread URL
  • Add Reddit link as comment on the original LinkedIn post (remind [Your Name])
  • Update website article frontmatter with all platform URLs:
    platforms:
      linkedin: "{url}"
      linkedin_technical: "{url}"
      x_thread: "{url}"
      x_thread_technical: "{url}"
      reddit: "{url}"
  1. Remind [Your Name] about engagement:

    "Article is live. Engagement windows: Reddit comments within 4 hours (critical), LinkedIn within 2 hours, X throughout the day."


Error Handling

  • Content agent failure at Stage 3/5: Retry once. If it fails again, write the draft directly (Jules writes it instead of the content agent).
  • Build failure at Stage 3e/5f: Fix the build error before proceeding. Common issues: frontmatter format, missing imports, image paths.
  • Posting failure at Stage 4c/6c/7c: Log the failure, continue with other platforms. Report all failures at the end. Manual posting is always the fallback.
  • LinkedIn token expired: Skip LinkedIn, note it in the tracking file. Remind [Your Name] to refresh: python3 Scripts/linkedin-auth.py

Resumability

If the pipeline is interrupted (session ends, context clears), it can be resumed:

  • Stage 2 complete? The decision spec in ~/.claude/plans/article-{slug}.md has everything needed.
  • Stage 3 in progress? The draft in Drafts/ is the checkpoint.
  • Stage 4 decision point reached? Check which platforms have been posted to. published.md tracks progress.
  • Stage 5-7 in progress? Same checkpoints apply.

To resume: "Continue /write-article for {title}" and Jules picks up from the last completed stage.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.