Postready

Post-Ready Pipeline

End-to-end content pipeline. Source material in → publish-ready LinkedIn posts out.

Runs four skills in sequence, automatically. No checkpoints. No menu. No asking which steps to run.

1. content — draft 3–4 posts from the source
2. factcheck — verify factual claims (Standard tier, no tier picker)
3. hatchet — tighten each post
4. ask-hr — HR-screen each post and apply rewrites for anything flagged

> Setup note: This skill orchestrates four other skills (content, factcheck, hatchet, ask-hr). All four must be installed. The path references below assume they live under your skills directory (e.g. ~/.claude/skills/content/SKILL.md) — adjust if yours are different.

---

Step 1: Generate the drafts (content)

Read the content SKILL.md and follow its instructions to:

• Identify and ingest the source content (pasted text, PDF, URL, email, newsletter, etc.)
• Find the best 3–4 angles
• Draft full-length posts in the user's voice (using the post skill)

Two important deviations from content's normal flow:

• Skip content's internal factcheck step. This pipeline runs factcheck centrally in Step 2 — don't double up.
• Do not "present" the posts yet. content normally ends by presenting finished drafts to the user. Here, treat the drafts as intermediate output and pass them to Step 2.

---

Step 2: Fact-check the drafts (factcheck — Standard tier)

Read the factcheck SKILL.md and run the Standard tier on every post that contains:

• Statistics or numbers
• Named figures, organizations, or surveys
• Specific historical, scientific, or attributed claims

Skip the tier picker entirely. Don't ask which tier — always Standard.

For each ❌ Incorrect or ⚠️ Partially correct finding, apply the suggested fix directly into the draft. Keep a one-line note per fix for the final summary (e.g. "Post 2: corrected '85% of jobs in 2030' → 'a widely cited but unsourced claim removed'").

If a post has no checkable claims, skip the search and note "no claims to check" in the pipeline notes.

---

Step 3: Tighten each post (hatchet)

Read the hatchet SKILL.md and run hatchet on each (now factchecked) post.

• Default target: roughly 50% reduction
• If a post is already tight, just clean it up — don't over-cut
• Preserve the user's voice and the sharp opening line
• Do not collapse the post into a single paragraph if the original used line breaks for rhythm

The hatchet output is intermediate — feed it into Step 4, don't present it yet.

---

Step 4: HR-screen each post (ask-hr)

Read the ask-hr SKILL.md and run an HR review on each tightened post.

For each post, decide based on the verdict:

| Verdict | Action | |---|---| | ✅ Clear or 🟢 Low Risk | Keep the post as-is from Step 3 | | 🟡 Moderate Risk | Use the Full Rewrite from ask-hr as the final version | | 🔴 High Risk | Use the Full Rewrite from ask-hr as the final version, flag prominently in pipeline notes |

Keep HR notes brief — name the issue and severity, one or two lines per flagged post. Don't include the full risk-score table in the final output.

---

Step 5: Present the final posts

Format the response exactly like this:

---

📦 Post-Ready: [N] posts

Post 1 — [angle type]

[Final post text — factchecked, tightened, HR-cleared]

---

Post 2 — [angle type]

[Final post text]

---

(repeat for each post)

---

🔧 Pipeline notes

Factcheck: [one line per fix applied, or "All claims verified" / "No checkable claims"]

Hatchet: [rough reduction summary, e.g. "Posts trimmed ~40–55%" or "All posts already tight, light cleanup only"]

HR: [any flags caught and the rewrite applied, or "All posts cleared"]

---

End by asking which (if any) the user wants to adjust further.

---

What NOT to do

• Do NOT pause for approval between steps — the whole point of postready is end-to-end automation
• Do NOT show intermediate drafts (raw, factchecked-but-not-trimmed, etc.) — only the final versions
• Do NOT skip any of the four steps on any post, even ones that look "obviously fine"
• Do NOT use a different factcheck tier — always Standard
• Do NOT over-cut tight posts in the hatchet step
• Do NOT open posts with newsletter meta-framing ("This week's newsletter…", "I read a PDF that…") — same rule as content
• Do NOT include the full ask-hr risk-score table or the full factcheck per-claim report in the final output — keep pipeline notes brief

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: analystacademy
• Source: analystacademy/claude-skills
• License: MIT

Install and usage instructions live in the source repository linked above.