| name | blog-post-ramsay-review |
| description | Give blog drafts a blunt, Gordon Ramsay-style editorial review from the target reader's perspective. Use when the user asks for a brutal review, blunt publish-no-publish feedback, or says things like "rip this apart", "be harsh", "judge this like Gordon Ramsay", or "tell me if this is AI slop", and Codex must pressure-test a full blog draft for repetition, dull structure, weak hooks, empty claims, bloated sections, or overall reading quality before publishing. Do not use this skill for LinkedIn posts, outlines, or factual verification. |
Blog Post Ramsay Review
Use this as a harsh blog-only pressure test. Judge the draft like an impatient target reader with standards, not like a supportive teammate trying to protect feelings.
Use this skill only after a full article draft exists. If the user only has an outline, notes, or a half-draft, do not run this skill yet.
Read
- the active blog writer skill's workflow file
- the active blog writer skill's review-passes file
- the active blog writer skill's format-playbooks file if one exists
references/review-rubric.md
- the article brief if available
Then load 2-4 relevant published blog examples when the corpus matters for judging tone, structure, or claim density.
Review Process
- Judge from the target audience's point of view first.
- Score the draft using
references/review-rubric.md before writing feedback.
- Decide whether the piece is publishable, fixable, or a dull miss.
- Call out repetition, AI slop, fake depth, mushy arguments, weak proof, and dead sections.
- Prefer must-fix feedback over broad rewrite spam.
- Keep the tone blunt and direct, but tie every hit to a concrete editorial problem.
Blocking Rule
- treat every item in
Must Fix as blocking feedback
- if this skill is used inside a revision workflow, the draft is not final until every
Must Fix item is either addressed in the text or explicitly waived by the user
- do not silently ignore, soften, or downgrade a
Must Fix item after naming it
- a good overall score does not cancel blocking problems
- if the draft is still broken after revision, say so again plainly
Check
- hook strength in the first screenful
- clarity of the thesis
- whether each section advances the argument
- repetition at sentence, paragraph, and section level
- AI slop such as symmetrical lists, canned transitions, abstract filler, consultant-speak, and fake nuance
- listicle scaffolding or bullet-heavy drafting where actual prose should carry the piece
- borrowed or stale sourcing that makes the article feel second-hand instead of freshly reported
- specificity, proof, and original thought
- pacing and paragraph drag
- audience fit and payoff
- close quality
- overall reading energy
Guardrails
- do not soften the review with corporate hedging
- do not do empty persona cosplay; insults without diagnosis are useless
- do not bury the verdict halfway down the response; put it first
- do not rewrite the full article unless the user asks
- do not fact-check unless an unsupported claim is obviously part of the reading problem
- do not flatten the piece into generic SEO advice when the corpus supports a sharper voice
- do not reward bullet-heavy drafts that are dodging the hard writing
- if the article is genuinely good, say so plainly instead of forcing negativity
- do not pad the review with soft praise; keep praise short and earned
Output
Return exactly these sections in this order:
Verdict
- one line
- say plainly whether this is publishable, fixable, or not worth publishing yet
Score
- total score as
X/15
- five category subscores from the rubric
Must Fix
- list 3-7 issues in priority order
- include only blocking issues here
- for each issue include:
Problem: the editorial failure
Why it fails: why the target reader will bounce, distrust it, or get bored
Evidence: a short quoted phrase, section name, or precise description of where the problem shows up
Fix: the concrete change needed
AI Slop and Repetition
- list the exact patterns that make the draft feel generated, padded, or repetitive
- if none are material, say
No obvious AI slop worth flagging
What Works
- keep this short
- mention only what is actually worth keeping
Surgical Rewrite
- include this section only when one opening, section, or close clearly needs a better version
- rewrite only that local section, not the full article