agentleFS
PingPingEPingPingE/ai-content-agents/.claude/skills/content-pipeline/SKILL.md

content-pipeline

Skill2 starsChanged 9 months agoMITView on GitHub

Full automated content pipeline — write, review, edit, and fact-check in one command

17 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. 0 are common, 15 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 11 edited versions we read keep these.

  • 15

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create, review, edit, and fact-check content in a single command.

    line 19
  2. Distinctive:

    If the brief is ambiguous, choose sensible defaults and state your assumptions at the top of the output

    line 34
  3. Distinctive:

    Use WebSearch to gather 5-8 credible sources on the topic

    line 38
  4. Distinctive:

    Create an outline with thesis, supporting points, and conclusion

    line 47
  5. Distinctive:

    Write the full draft following the template structure

    line 48
  6. Distinctive:

    Verify word count falls within the Word Count Guidelines for the content type

    line 51
  7. Distinctive:

    If the review verdict is REVISE, or after a REJECT rewrite:

    line 64
  8. Very specific:

    After editing, re-submit to content-reviewer for a second review

    line 67
  9. Distinctive:

    If the second review is still REVISE or REJECT:

    line 68
  10. Distinctive:

    Do NOT request a third review — proceed to Phase 6

    line 70
  11. Distinctive:

    If the review is APPROVE — proceed to Phase 6

    line 71
  12. Distinctive:

    Run a final self-review against the quality checklist:

    line 85
  13. Distinctive:

    If content is still not APPROVE after 2 cycles, finalize with the best version and note remaining issues in the quality report

    line 98
  14. Distinctive:

    Follow all phases in order — do not skip Research or Fact-Check

    line 167
  15. Distinctive:

    Use Task to delegate to other agents — never impersonate them

    line 170
  16. Distinctive:

    If WebSearch fails or returns no results, note the gap and proceed with available information

    line 172

“Common” is measured: kept by at least 6 in 10 of the 11 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

Content Pipeline — End-to-End Automated Content Creation

Create, review, edit, and fact-check content in a single command.

Brief: $ARGUMENTS

Instructions

This skill chains all four content agents into one automated pipeline. Follow each phase in order. Cap revision cycles at 2 to prevent infinite loops.

Phase 1: Parse the Brief

  1. Extract from $ARGUMENTS:
    • Topic — the subject to write about
    • Content type — blog post, tutorial, API docs, or landing page (default: blog post)
    • Target audience — developers, executives, end users, or marketers (default: developers)
    • Special requirements — tone, word count, keywords, or other constraints
  2. If the brief is ambiguous, choose sensible defaults and state your assumptions at the top of the output

Phase 2: Research

  1. Use WebSearch to gather 5-8 credible sources on the topic
  2. Prioritize primary sources: official documentation, research papers, industry reports
  3. Extract key data points: statistics, quotes, examples, and counterarguments
  4. Identify the unique angle — what do existing articles miss?
  5. Record all source URLs for citation in the draft

Phase 3: Draft

  1. Select the matching Content Template from your agent guidelines (blog post, tutorial, API docs, or landing page)
  2. Create an outline with thesis, supporting points, and conclusion
  3. Write the full draft following the template structure
  4. Include inline citations for all facts and statistics
  5. Save the draft to a file (e.g., content/<slug>.md)
  6. Verify word count falls within the Word Count Guidelines for the content type

Phase 4: Review

  1. Delegate to content-reviewer via Task:
    • "Review the content at content/<slug>.md using the full 6-dimension scoring rubric. Return the scored report with verdict (APPROVE/REVISE/REJECT) and action items."
  2. Parse the reviewer's response:
    • APPROVE (score >= 24/30) — proceed to Phase 6 (Fact-Check)
    • REVISE (score 18-23/30) — proceed to Phase 5 (Edit)
    • REJECT (score < 18/30) — rewrite the draft addressing all action items, then re-submit for review (counts as revision cycle 1)

Phase 5: Edit

  1. If the review verdict is REVISE, or after a REJECT rewrite:
    • Delegate to editor via Task:
      • "Edit the content at content/<slug>.md. Apply all three editing passes (structure, clarity, grammar). Target 20% word reduction. Address these reviewer action items: [paste action items]."
    • After editing, re-submit to content-reviewer for a second review
  2. If the second review is still REVISE or REJECT:
    • Apply the remaining action items yourself (this is revision cycle 2)
    • Do NOT request a third review — proceed to Phase 6
  3. If the review is APPROVE — proceed to Phase 6

Phase 6: Fact-Check

  1. Delegate to fact-checker via Task:
    • "Fact-check the content at content/<slug>.md. Verify every statistic, quote, technical claim, and external reference. Return the full claim verification table."
  2. Parse the fact-checker's report:
    • Incorrect claims — fix immediately with the correct information and source
    • Unverified claims — add a source if one exists, or soften the language ("approximately", "reportedly")
    • Verified claims — no action needed

Phase 7: Final Polish

  1. Apply all fact-check corrections to the content file
  2. Run a final self-review against the quality checklist:
    • All facts cited with sources
    • Headings descriptive and scannable
    • Paragraphs 3-4 sentences max
    • No weasel words or filler phrases
    • Clear call-to-action or next steps
    • Word count within guidelines
    • Reading level appropriate for audience
  3. Save the final version

Revision Cycle Limits

  • Maximum 2 revision cycles through Phase 4-5
  • If content is still not APPROVE after 2 cycles, finalize with the best version and note remaining issues in the quality report
  • This prevents infinite review-edit loops while still producing high-quality content

Output Format

Deliver two artifacts:

1. Final Content File

Saved to content/<slug>.md with metadata footer:

# [Title]

[Final content here]

---

## Metadata
- **Word count:** [number]
- **Reading time:** [X minutes at 250 words/min]
- **Content type:** [blog post / tutorial / API docs / landing page]
- **Target audience:** [audience]
- **Primary keyword:** [keyword]
- **Sources used:** [count]

2. Quality Report

Print to the conversation:

## Content Pipeline Report

### Pipeline Summary
| Phase | Status | Notes |
|-------|--------|-------|
| Research | DONE | [X] sources gathered |
| Draft | DONE | [word count] words |
| Review | [APPROVE/REVISE] | Score: [X]/30 |
| Edit | [DONE/SKIPPED] | [X]% word reduction |
| Fact-Check | DONE | [X] verified, [Y] corrected, [Z] unverified |
| Final Polish | DONE | [corrections applied] |

### Revision History
- **Cycle 1:** [summary of changes]
- **Cycle 2:** [summary or "Not needed"]

### Review Scores (Final)
| Dimension | Score |
|-----------|-------|
| Accuracy | X/5 |
| Clarity | X/5 |
| Style | X/5 |
| Structure | X/5 |
| Completeness | X/5 |
| Audience | X/5 |
| **Total** | **X/30** |

### Fact-Check Summary
- **Total claims:** [X]
- **Verified:** [Y]
- **Corrected:** [Z]
- **Unverified:** [W]

### File Location
`content/<slug>.md`

Rules

  • Follow all phases in order — do not skip Research or Fact-Check
  • Maximum 2 revision cycles, then finalize
  • Save the content file after each phase so progress is preserved
  • Use Task to delegate to other agents — never impersonate them
  • Cite every factual claim with a source URL
  • If WebSearch fails or returns no results, note the gap and proceed with available information

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.