ai-content-quality-gate
Skill1 starsChanged 23 days agoMITView on GitHub
Reviews Artificial Intelligence (AI)-assisted marketing content before publication and returns a publish / revise / kill verdict with an AI-tell detection checklist, a brand-voice score, a factual-claim verification log, an originality and added-value test, and line-level fixes. Use when reviewing a blog post, landing page, email, or social post drafted with an AI writing tool; setting up an editorial review step for AI-generated content at scale; content sounds generic or machine-written; or claims and statistics in a draft need verification before publishing.
In this repository41 other files it gives its agents.
- ai-activation-optimizerSkill
- ai-feature-experiment-designerSkill
- ai-launch-playbookSkill
- ai-pricing-margin-modelerSkill
- ai-product-positioningSkill
- ai-search-visibility-optimizerSkill
- agent-autonomy-level-mapperSkill
- ai-failure-state-designerSkill
- ai-usability-test-plannerSkill
- ai-ux-auditSkill
- human-in-the-loop-designerSkill
- prompt-surface-designerSkill
- uncertainty-and-explainability-patternsSkill
- ai-agent-financial-analystSkill
- burn-rate-runway-plannerSkill
- capital-allocation-frameworkSkill
- dcf-valuation-builderSkill
- financial-statement-analyzerSkill
- saas-cohort-analyzerSkill
- competitive-response-modelerSkill
- mechanism-design-plannerSkill
- nash-equilibrium-solverSkill
- negotiation-strategistSkill
- repeated-game-strategistSkill
- ai-decision-engineSkill
- ai-product-strategySkill
- ai-stakeholder-translatorSkill
- customer-interview-synthesizerSkill
- jobs-to-be-done-extractorSkill
- opportunity-solution-treeSkill
- product-discovery-coachSkill
- product-roadmap-prioritizerSkill
- blue-ocean-strategy-canvasSkill
- go-to-market-strategySkill
- gtm-engineeringSkill
- market-sizingSkill
- okr-cascade-plannerSkill
- porters-five-forces-analyzerSkill
- pricing-strategy-analyzerSkill
- strategic-landscape-mapperSkill
- swot-tows-analyzerSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 8 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
Search and social content penalized or ignored for low added value
- Distinctive: line 25
Before publishing any AI-assisted blog post, landing page, email, or social post
- Distinctive: line 47
Ask: could a reader get this answer by asking a general-purpose assistant the headline as a question?
- Distinctive: line 94
Call to action matches the stated goal
- Distinctive: line 105
Every Revise verdict includes: the top 3 changes that most improve the piece, line-level rewrites for flagged tells, and the claims to fix …
- Distinctive: line 119
Verify facts before judging style
- Distinctive: line 120
Ask for a brand-voice reference before scoring voice
- Distinctive: line 127
Treat removal of AI tells as sufficient when the piece has no added value
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI Content Quality Gate
AI makes drafting free - this skill makes sure what ships is still worth a reader's time and safe to put your name on.
What this skill is
A pre-publication review workflow for marketing content produced with help from Large Language Model (LLM) writing tools. It checks five things in order: whether the piece is factually safe, whether it adds something a reader could not get from asking an assistant directly, whether it sounds like the brand rather than a model, whether it is structurally sound, and whether it meets channel requirements. It ends in a clear verdict with specific fixes, so editorial review becomes a consistent gate instead of a matter of taste.
What it solves
- Published statistics, quotes, or case details that were invented by a model
- Content that is fluent but says nothing a reader could not get from a general-purpose assistant
- Recognizable AI tells that erode reader trust and brand distinctiveness
- Inconsistent editorial standards across writers, agencies, and AI tools
- Search and social content penalized or ignored for low added value
- Review feedback like "make it pop" instead of actionable line edits
When to invoke
- Before publishing any AI-assisted blog post, landing page, email, or social post
- Setting a standard review step for a team producing AI-drafted content at volume
- Auditing a backlog of existing content for quality and accuracy
- A draft feels generic, padded, or obviously machine-written
- Content includes statistics, customer claims, or competitor comparisons
Phase 1: Intake
Record: content type, channel, target reader, single goal (inform, convert, retain), primary claim, brand voice guide (or 3 approved reference pieces), and which parts were AI-drafted. Without a voice reference, Phase 4 scoring is not reliable - ask for one.
Phase 2: Factual-claim verification log
Extract every checkable claim and verify before anything else. A single fabricated fact is an automatic block.
| Claim | Type | Source cited | Verified? | Source quality | Action |
| Statistic / quote / customer result / product capability / competitor statement / date | Yes / No / Partial | Primary / secondary / none | Keep / correct / cite / cut |
Rules: statistics need a named primary source and year; quotes need a confirmed speaker; product capabilities must match the current release; competitor statements must be verifiable public facts; customer results need approval.
Phase 3: Originality and added-value test
Ask: could a reader get this answer by asking a general-purpose assistant the headline as a question? Score each 0-2:
| Added-value signal | 0 | 1 | 2 |
| First-party data or results | None | Referenced | Central to the piece |
| Expert experience | Generic | Some specific examples | Named practitioner insight, lessons learned |
| Point of view | Balanced summary | Mild stance | Clear, argued position |
| Specificity | Abstract advice | Some examples | Concrete steps, numbers, templates |
| Unique asset | None | Minor (quote, chart) | Framework, tool, dataset, or case study |
Added-value score = sum of the five signals (0-10)Under 4 is a commodity piece; it is a kill or a full rewrite regardless of polish.
Phase 4: AI-tell detection and brand-voice score
AI-tell checklist (count occurrences, flag lines):
| Tell | Example pattern |
| Inflated openers | "In today's fast-paced world", "In the ever-evolving landscape" |
| Stock vocabulary | delve, tapestry, unlock, elevate, seamless, robust, game-changer, navigate the complexities |
| Rule-of-three padding | Lists of three adjectives or clauses in most paragraphs |
| Contrast formula | "It's not just X, it's Y" repeated |
| Hedged non-claims | "can potentially help", "may play a role" |
| Uniform rhythm | Every paragraph the same length; every section ends with a summary line |
| Generic conclusions | "In conclusion", restating the introduction |
| Unsourced authority | "Studies show", "experts agree" |
| Emoji or heading overuse | Decorative structure with thin content |
Brand-voice score, 1-5 per dimension against the voice guide:
| Dimension | 1 | 5 |
| Vocabulary | Generic, model-default words | Brand's own terms and phrasing |
| Tone | Mismatched (too formal or too casual) | Matches reference pieces |
| Sentence rhythm | Uniform | Varied, natural |
| Perspective | Faceless | Recognizable point of view |
| Reader fit | Written for everyone | Written for the stated reader |
Phase 5: Structure and channel checks
- Headline promise is delivered in the body
- The first 2 sentences state the point; no throat-clearing
- One idea per section; every section earns its place
- Call to action matches the stated goal
- Channel requirements: length, links, accessibility (alt text, descriptive links), legal disclaimers, required AI-use disclosure where policy or regulation calls for it
Phase 6: Verdict
| Verdict | Criteria |
| Publish | All claims verified, added-value 7+, voice average 4+, 3 or fewer AI tells, structure and channel checks pass |
| Revise | All claims verified or fixable, added-value 4-6 or voice average 3-3.9 or 4+ AI tells; return with line edits |
| Kill | Any unverifiable material claim that the piece depends on, added-value under 4, or the piece duplicates existing content |
Every Revise verdict includes: the top 3 changes that most improve the piece, line-level rewrites for flagged tells, and the claims to fix or remove.
Output
- Verdict (Publish / Revise / Kill) with the criteria that drove it
- Factual-claim verification log
- Added-value score with the weakest signals named
- AI-tell list with line references and suggested rewrites
- Brand-voice scorecard
- Prioritized revision list (top 3 changes first)
Operating rules
Always
- Verify facts before judging style
- Ask for a brand-voice reference before scoring voice
- Give line-level fixes, not general impressions
- Judge whether the piece adds value beyond what an assistant would answer directly
- Apply the same thresholds to human-written and AI-assisted drafts
Never
- Pass content containing an unverified statistic, quote, or customer result
- Treat removal of AI tells as sufficient when the piece has no added value
- Fabricate replacement sources or data when a claim fails verification
- Rely on AI-detection tools as proof of authorship
- Publish competitor comparisons that cannot be backed by public, dated evidence
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
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.

