Loading the tool…
The swap test
Put someone else's name on your post. Did anything break?
That is the whole test. If the post survives a byline swap without losing anything, it was never yours to begin with, and the reach it earns builds somebody else's reputation as easily as yours. Paste a post below and twelve checks run on it in your browser.
Paste and run · nothing stored · no account
Most people get this completely wrong.
Leadership isn't about having all the answers. It's about asking better questions.
I've learned that the best leaders are curious, humble, and consistent. They show up when it's hard. They listen more than they speak.
The lesson? Trust is built in the small moments.
Thoughts?
Nothing broke. That is the problem.
Why it matters at your level
Generic content is expensive in proportion to your standing
A junior professional publishing something forgettable loses nothing, because nobody had expectations. A senior one publishing the same thing spends credibility they took twenty years to accumulate, in front of an audience that knows what they are capable of.
Cost one
Your peers can tell
The people whose opinion of you actually matters are the ones best equipped to spot content that could have come from anywhere. They will not comment on it. They will simply revise their estimate of how much you have left to say.
Cost two
Platforms learn what you are
Ranking systems match content to interest by learning what an account reliably produces. Content indistinguishable from ten thousand other accounts gives the system nothing specific to learn, so it gets matched broadly and lands nowhere in particular.
Cost three
You cannot compound what nobody attributes to you
Reputation accrues to a recognizable voice. If ten good posts read as ten different people wrote them, you have published ten times and built nothing, which is the most expensive possible outcome for the hours spent.
The reference list
Twelve tells, in four families, weighted by how much they cost you
The first three families are surface. They are habits, they are fixable in one editing pass, and they are what most people mean when they say content reads as machine-made. The fourth family is the one that decides the test, and no amount of rhythm work compensates for failing it.
Rhythm
19 points · the cadence that gives it awayThe dash lattice
7 pointsLong dashes used as a rhythm device. One is a choice. Two or more in a short post is a fingerprint, and it is currently the most recognized tell in professional writing.
BeforeThe work was hard — harder than expected — and we shipped anyway.
AfterThe work was hard. Harder than expected. We shipped anyway.
AfterThe work was hard. Harder than expected. We shipped anyway.
The triad
7 pointsThree parallel items where two would do and one would be braver. The pattern reads as balanced, which is why it is everywhere, and balance is not a thing anyone remembers you for.
BeforeCurious, humble, and consistent.
AfterCurious enough to be annoying about it.
AfterCurious enough to be annoying about it.
The staccato run
5 pointsThree or more very short sentences in a row, used for emphasis. It reads as urgency at first and as a format shortly afterwards.
BeforeWe shipped. It failed. We learned.
AfterWe shipped it in March and pulled it in May, which cost us a quarter and one very patient customer.
AfterWe shipped it in March and pulled it in May, which cost us a quarter and one very patient customer.
Structure
28 points · borrowed shapesThe negation reframe
9 pointsSetting up a straw position and pivoting off it. The construction manufactures the appearance of insight by inventing something to correct, and the correction is usually something everyone already believed.
BeforeIt's not about the tools. It's about the process.
AfterWe bought the tool first and spent four months discovering the process was the problem.
AfterWe bought the tool first and spent four months discovering the process was the problem.
The setup opener
7 pointsAn opening line that withholds rather than delivers. It promises a payoff and buys a second of attention against the reader's expectation, which is a loan taken out on your credibility.
BeforeMost people get hiring completely wrong.
AfterWe hired four sales leaders in three years. Three of them were the same mistake.
AfterWe hired four sales leaders in three years. Three of them were the same mistake.
The question hook
5 pointsOpening on a question hands the reader an exit. A statement they disagree with keeps them reading to argue. A question they can answer in their head lets them leave having already participated.
BeforeWhat separates good leaders from great ones?
AfterGreat leaders are worse at delegating than good ones, and it costs them.
AfterGreat leaders are worse at delegating than good ones, and it costs them.
The lesson landing
7 pointsEnding by extracting a moral. It converts a specific story into a general platitude in the final line, which is backwards, since the specific part was the only part worth reading.
BeforeThe lesson? Trust is built in small moments.
AfterShe never mentioned it again, and I have thought about it every quarter since.
AfterShe never mentioned it again, and I have thought about it every quarter since.
Vocabulary
19 points · the hollow lexiconHollow words
8 pointsWords that sound like meaning and carry none. They survive drafting because they feel professional, and they read to a senior audience as filler.
BeforeWe leveraged our learnings to unlock growth at scale.
AfterWe stopped doing outbound and revenue went up.
AfterWe stopped doing outbound and revenue went up.
The announcement opener
6 pointsHumbled to announce. Thrilled to share. The register is so standardized that readers skip the line automatically, which means you spent the best position in the post on words nobody read.
BeforeHumbled and thrilled to announce I've joined Acme as VP.
AfterI start at Acme on Monday. Here is the problem I took the job to fix.
AfterI start at Acme on Monday. Here is the problem I took the job to fix.
The engagement plea
5 pointsAsking for the engagement announces that you did not expect the content to earn it, and senior readers read the request exactly that way.
BeforeAgree? Let me know in the comments.
AfterNothing. End on the last real sentence and let it sit there.
AfterNothing. End on the last real sentence and let it sit there.
Substance
34 points · the family that decides itThe specificity void
18 pointsNo numbers, no names, no places, no dates. Nothing a reader could check, dispute, or remember. This outranks every surface tell, because a post with real specifics survives clumsy rhythm and a post without them cannot be rescued by good rhythm.
BeforeWe saw significant improvement after the change.
AfterResponse time went from nine days to two, and the complaints stopped in about a month.
AfterResponse time went from nine days to two, and the complaints stopped in about a month.
The missing incident
16 pointsNothing happened. No moment, no decision, no consequence. General advice about a topic is available from anyone with the same job title, and the thing you have that they do not is the specific afternoon it went wrong for you.
BeforeHiring senior people requires patience and clarity.
AfterI offered the job on a Thursday. By Monday he had taken a counter, and I had learned to stop negotiating on Thursdays.
AfterI offered the job on a Thursday. By Monday he had taken a counter, and I had learned to stop negotiating on Thursdays.
Run the test
Paste a post you have already published
Something real, ideally something you were reasonably pleased with. The checks run in your browser and the text never leaves this page.
Needs at least 40 words. Below that the checks produce noise rather than a reading, and a tool that returns a confident verdict on two sentences is lying to you.
The honest part
One post takes five minutes. Every post, forever, is the actual job.
Nothing here is difficult once you can see it. The problem is that seeing it is a skill you have to apply on your own drafts, when you are tired of them, at the exact moment you want to press publish and stop thinking about it. The tells creep back within a month of anyone deciding they have got the hang of this.
Doing it by hand
What holding the line costs
Running the check on a draft~5 min / post
Rewriting what fails, which is most of them at first~20 min / post
Finding the specific detail the post was missing~10 min / post
Re-reading it cold the next morning~5 min / post
Across a 40-post quarter~26 hours of editing
Agent Craft's own time estimate, not a published statistic, and it assumes the writing itself is already done. You become your own editor, indefinitely. Most people manage it for about four weeks, which is roughly when the standard quietly drops back to whatever comes out first.
Agent Craft
Every draft passes this screen before you see it
Content that could have come from anyone is our definition of a defect, so it is caught at the draft stage rather than after publishing. You record a two-minute voice note. What comes back has the specifics in it, because the specifics came out of your mouth.
- The twelve checks run on every draft automatically, and drafts that fail get rewritten before they reach you
- Your voice, your banned words and your positioning held as standing rules rather than something you police manually
- One voice note re-formed per platform across LinkedIn, X, TikTok, and YouTube, since a post syndicated unchanged reads exactly like a post syndicated unchanged
- Strategy included, guided by a marketing lead with twenty years running global brands
30-day free trial · no credit card required
Before you ask
Reasonable questions
Does this detect whether a machine wrote my post?
No, and any tool claiming to do that reliably is overselling. This checks whether the writing is distinctive, which is a different and more useful question. A person can write something entirely generic without any help, and it costs them exactly as much either way.
Is my post sent to a server?
No. The twelve checks are pattern rules running in your browser, which is why the result is instant and why we can promise this. When you submit your email, four numbers go with it: your score, the count of tells found, and the counts by family. The text itself stays where you pasted it.
I scored low on a post I am proud of. Is the tool wrong?
Possibly. The checks are pattern rules, so they cannot read intent, and a deliberate stylistic choice looks identical to a habit from the outside. Read the specific findings rather than the number. If a flag names something you did on purpose and would defend, you have your answer and the score is noise.
Why do the substance checks count for a third of the score?
Because they are the only two that cannot be fixed with editing. A post with no specifics and no incident in it has nothing wrong with its sentences. It has nothing in it, and rewriting the rhythm produces a smoother version of the same problem.
Which habits does it under-report rather than over-report?
Several, deliberately. The triad only fires on a list written with a serial comma, so “curious, humble and consistent” passes. A moral extracted in the middle of a post rather than its final lines passes. And no pattern rule can tell a deliberate choice from a habit. Every one of those errs toward saying nothing rather than toward accusing you, which is the right direction for a tool that scores somebody's writing.
Who is this not for?
Anyone who does not publish, since there is nothing to paste. Anyone working from templates, because the tool will score their entire approach honestly and the result reads as an indictment. And anyone wanting reassurance, since the checks flatter nobody, including us: this page was written under the same twelve rules.