Writing bullets that survive a sceptical reader
Duty phrasing, missing evidence and repeated verbs account for most of what makes a resume forgettable. All three are fixable without inventing anything.
The sentence that costs you the most
"Responsible for the CI/CD pipeline" describes the job you were handed. It says nothing about what you did with it, and two people responsible for the same thing can have done entirely different work. The reader cannot tell which one you were, so they assume the less interesting answer.
The same is true of "worked on", "assisted with", "involved in" and "part of a team that". Each of them is a way of naming a responsibility without claiming a result, and resumes are full of them because they are safe. Safe here means invisible.
The fix is not a stronger verb. It is answering the question the phrase avoids: what did you actually do, and what was different afterwards.
Before: Responsible for the CI/CD pipeline.
After: Rebuilt the release pipeline and took deploy time from 45 minutes to 8, which moved the team from weekly releases to daily.
The second version is longer, and it is worth the words because it contains three things a reader can hold on to: what changed, by how much, and what it enabled.
Numbers are not decoration
A bullet without a figure asks the reader to take your word for the size of it. A bullet with one lets them compare you against the last candidate they read about, which is what they are actually doing.
Useful figures are not only money and percentages. Scale counts: forty services, twelve people, three countries, two thousand tickets a month. So does duration: a two-week turnaround, a three-year run, a migration completed in one quarter. So does a comparison: from forty-five minutes to eight, ahead of a deadline, under a budget.
Where you genuinely do not have the number, do not estimate one. Say what changed in plain words instead. An invented figure is the first thing an interviewer asks you to expand on, and it is a bad conversation to have. A resume you cannot defend in a room is worse than a thin one.
Repeating one verb flattens everything
Six bullets that all open with "Led" read as one bullet. The eye stops distinguishing them and the achievements blur into a single impression of leading things.
The fix is not a thesaurus. Reaching for "spearheaded" because "led" appeared three times already produces the register that makes resumes unreadable, and it quietly changes the claim — "spearheaded" implies initiating, and you may only have run something that already existed.
The bullet and verb checker offers alternatives drawn only from verbs meaning roughly the same thing, grouped by what the action actually is: built something new, improved something existing, made something smaller, made something bigger, ran something ongoing. Swapping within a group keeps your claim identical. Changing the claim is a decision for you, not for a tool.
Three orders, three different claims
The same facts can be arranged several ways, and the arrangement is itself a claim.
Action first — "Cut cloud spend across 40 services, saving $240k a year" — is the safe default. It says what you did, then what followed, which is how a reader expects a bullet to run.
Result first — "Saved $240k a year by cutting cloud spend across 40 services" — is stronger when the figure is the point. But it asserts that the outcome followed from your action, which is a bigger claim than it looks. Only use it where that is genuinely true.
Method named — "Cut cloud spend across 40 services by rightsizing instances, saving $240k a year" — earns its extra words when the method is the skill the job asks for. Otherwise it is detail nobody needs.
The achievement statement generator produces all three from parts you supply, and tells you what each ordering claims. It adds no word you did not type.
A skills list is a weak signal
Everyone knows a skills list can be written in a minute, so everyone discounts it. Twelve terms with nothing behind them are worth less than four that appear in something you describe doing.
The useful test: for each term in your list, can you point at a line elsewhere in the resume where you used it? If not, either add that line or cut the term. The skills extractor does exactly this comparison and separates the two groups.
The reverse case is more common than people expect and costs nothing to fix. If your experience describes running a Postgres estate and your skills line does not say Postgres, a keyword search over the list misses you entirely.
LinkedIn is read in a list, not on your profile
Most people who see your headline are scanning search results or a queue of connection requests. They read the first line and decide whether to click. The field stores 220 characters and displays far fewer — around 120 is what reliably survives.
That is why the plain arrangement, role then what you are good at, usually beats the clever one. It survives truncation and reads the same in every context. The headline tool counts each arrangement against both limits.
The About section has the opposite problem: it holds 2,600 characters and shows roughly the first 300 before a "see more" link. Everything after the fold is read by people who have already decided you are interesting, so put the thing that makes them decide that above it. The About builder counts the visible part separately.
What none of these tools will do
None of them writes a sentence for you. Where something is missing you get the question — what did you do, how much, compared with what — because the answer is a fact only you have.
That is a deliberate limit rather than an unfinished feature. A generator that fills the gaps produces bullets you cannot defend, and the place you find that out is an interview.