// Strategic content production system — writer context packet, bookmarkability rubric, avoid-slop patterns, and viral postmortem. Load when planning, drafting, scoring, or reviewing any post.
| name | content-system |
| description | Strategic content production system — writer context packet, bookmarkability rubric, avoid-slop patterns, and viral postmortem. Load when planning, drafting, scoring, or reviewing any post. |
| metadata | {"tags":"content-strategy, editorial, quality-gates, slop-detection, rubric"} |
A structured content production system adapted from Shann Holmberg's framework (5M impressions in 2 weeks). Use this skill when planning, drafting, scoring, or reviewing content for the site.
The unit of work is not "another post." It is something a reader wants to keep. Content is only as good as the person and the system behind it. The agent count is not the lever; the knowledge layer feeding the writer is.
Every piece of content follows one path: idea → brief → draft → verify → publish → feedback → archive.
Each stage has a defined artifact and a clear handoff.
inbox → idea.md → brief.md → draft-package.md → [verify] → publish → feedback.md → archive
A content object lives in its own run folder until it ships. One piece of content = one folder. The folder carries its own state.
The system depends on six directories. Map these to your existing project structure rather than creating parallel infrastructure.
/strategy — positioning, audience, pillars, source watchlist
/voice — voice profile, avoid-slop patterns
/stores — inbox, workboard, ideas, hooks, proof, feedback
/runs/active — one folder per in-flight content object
/runs/archive — shipped content objects
/modules — skill files, references, templates
/workflows — idea-to-publish steps, verifier, scheduler, feedback loop
| System directory | This project equivalent |
|---|---|
/strategy | CLAUDE.md (positioning, audience, editorial rules) |
/voice | CLAUDE.md "Writing Voice & Editorial" section + essay-authoring skill voice guardrails |
/stores/proof | agents/ (running agent code), agent-log.json (production telemetry), blog posts with real data |
/stores/inbox | Notion: Proactive Agents / The genesis (briefs, ideas, status) |
/runs/active | Notion post pages with status tracking |
/modules | essay-authoring/, content-system/ (skill directories at repo root) |
/workflows | essay-authoring skill (full authoring workflow) |
Before drafting any content, build a tight brief. A tight packet beats a giant context window. The brief should be 400-900 tokens.
thesis: one sentence the post must prove
reader: the specific person who should save it
proof: numbers, screenshots, stories we are allowed to use
angle: the unexpected framing
constraints: format, length, tone, banned phrases
voice anchors: 2-3 lines that sound like us
risks: what would make this read as slop
open loops: what we do not yet know, that the writer should flag
Score every draft before publishing. 0, 1, or 2 points each. Our bar is 8 out of 12.
| Row | Question |
|---|---|
| 1 | Does it save the reader a future task? |
| 2 | Does it include proof (numbers, named example, screenshot)? |
| 3 | Does it give a reusable takeaway (template, checklist, framework)? |
| 4 | Does it have a specific audience and job-to-be-done? |
| 5 | Can it be applied without us being in the room? |
| 6 | Does it have a strong visual (figure, table, diagram)? |
Below 8: the draft goes back to the brief, not to the trash. Most weak drafts are good drafts that skipped one row in the rubric. Fix the row, re-score, ship.
These patterns mark content as AI-generated. Run every draft against this list. Organized by severity.
Before publishing, run this analysis on the draft. The model must point at exact lines, not offer generic praise.
For each category, identify a specific line:
If you cannot point at a specific line for any category, that category needs work before publishing.
Writing and orchestration are different jobs. They reward different approaches.
Writer concerns: taste, rhythm, compression, voice, the actual draft. Use the strongest available model. Prioritize voice fidelity over speed.
Orchestrator concerns: routing between layers, packaging context for the writer, deciding what gets passed in, running the verifier, handoff to publish. Prioritize reliability and tool access over prose quality.
In practice for this project: the essay-authoring skill handles the writer role, this content-system skill handles the orchestration and quality control role.
After every published piece:
The feedback loop is the moat. A system without feedback produces the same quality forever. A system with feedback compounds.