Six stations, one Career File, one operating system for a job search
Most job searches fail on process, not effort. Hours go into scrolling boards that resurface the same closed roles, and then more hours go into tailoring resumes for postings nobody intends to fill. Postcheck splits that work into six stations and puts a rule behind each one.
Nothing in the guide matches that. Try one word rather than a sentence.
Six stations on one line, and one record underneath all of them. None of them applies for you.
| Station | The question it answers | What comes out |
|---|---|---|
| Board | What should I do next? | Due decisions, applications ready, contacts to make, interviews coming up, follow ups due |
| Scout | What is open that clears my gates and scores above my threshold? | A prompt or a plan file that produces a ranked daily shortlist |
| Decide | Does this posting earn an hour of my tailoring time? | One call per role: Apply, Investigate First, or Pass |
| Apply | Will the scanner read this, and does my language match theirs? | One call on the document: Send as is, Fix first, or Rewrite |
| Connect | What do I actually say to a human? | A message you can send, and a record of when you sent it |
| Prepare | Can I answer what they will ask with a number attached? | A prep sheet, the questions to ask back, a debrief, and an offer score |
Underneath the six sits a seventh screen that is not a station: your career profile. It is where the whole record is set, and every station reads from it. The app calls that record your Career File, and the bar across the top of every screen is where you back it up, restore it, export the board, or erase everything.
Open your career profile in the appOpens in a new tab so this guide stays where it is.
Everything the app knows about you lives in one record. Your name and contact details, your one line summary, your proof points, your resume library, your ranked target titles, your ranked industries, your geography, your compensation, your deal breakers, and your sources. Every station reads from it. Nothing you enter is uploaded and there is no account. The page you are reading loads Google Analytics and Google Fonts, so anonymous counts of which pages get opened reach Google, and nothing from your Career File goes with them.
Fill it before anything else. A bar across the top of the profile counts what is still blank and offers to take you to the next empty field. The scout reasons from these fields, so a blank salary floor produces a scout with nothing to enforce and a blank exclusion list produces one that will happily hand you a role at an organization you refuse to work for. On a first visit with nothing saved, the app opens here rather than on the Board, for exactly that reason.
Every version you send is a row: a name, the text, and where the text came from. Add one as a PDF or paste it in. You can also add your LinkedIn profile as a source rather than a resume. One row is in use at a time. The one in use is what filled these profile fields, it is what the Apply station checks every posting against, and the Board reports which version is actually getting replies.
Read what the import filled in. It reads the text of a PDF and makes its best guess at your titles and proof points. It is a head start, not a transcript. Correct it before you move on, because whatever sits in those fields is what the scout reasons from. The gentle fill never overwrites something you typed yourself, which is why switching resumes can look like nothing happened. Every version carries an explicit replace action for when you do want it to overwrite.
Print a clean copy. Most of what the Apply station flags is layout, not writing: headings a scanner cannot read, contact details stranded in a footer, two columns, a table. One button rebuilds the structure a scanner reads, with headings named plainly, the contact block in the body, one column and plain bullets, and it does not change a single word you wrote. Print it, save it as a PDF, then add that file back as a version. The file you send is then the file the app checked.
Titles and industries are ranked lists. Number one is the job you want most, and the scout leads with the top of the list. Type your own or take a suggestion from the vocabulary. Nothing you type is ever refused or rewritten, and a title that is not on the list travels like any other.
Geography is a county list plus a distance. Name the counties, because without a list the scout invents a commute radius. Then, if you want one, add your home ZIP code and a maximum one way distance. Give both and the scout uses both. Leave the distance blank and it uses the county list alone.
Compensation is structured. A salary type, whether annual, hourly or contract. A desired range with a low and a high. A minimum acceptable. A hard floor. The floor is a hard gate and nothing below it is ever surfaced. The range and the minimum are scoring inputs. There is also a setting for postings with no salary line at all: show them, show them and flag them to verify, or do not show them. That setting is written into the scout prompt, so it changes what the scout is told rather than sitting in the app as a preference nobody acts on.
Deal breakers and gray areas. Organizations to never surface, industries to never surface, and a separate line for gray areas you want flagged for your review rather than dropped. An excluded organization may never appear in a report as a find, as a source, or in a reference list.
Where to search. Employer career pages beat job boards, and the scout opens three of them before it touches a board. If you are not sure who belongs on that list, one button writes a prompt that finds employers from your industries and your geography. The app sends nothing anywhere. You copy the prompt, paste it into an assistant yourself, and paste the answer back.
Job site chips. Your job sites field carries twenty one one click chips: the national boards, the nonprofit boards, and the Texas boards. Tap what you want rather than typing it, or select all and cut it back.
Backing up is one button and one file. Everything lives in the browser on the device you are using. It does not sync. Browser storage does get cleared, and a cleared browser takes the whole pipeline with it, so the app nags you every twenty saves. The backup is a plain text file. Windows may ask which app to open it with, and some mail systems hand it back named .txt. Restore takes it either way. Do not rename it and do not edit what is inside.
Install it on your phone. Android and desktop Chrome offer to install it. On iPhone, open it in Safari, tap Share, then Add to Home Screen. Once installed it opens with no browser bar and keeps working with no signal, which is what makes it usable in a waiting room or on a bus.
Open the Board in the appOpens in a new tab so this guide stays where it is.
The Board is the front door and the tracker. It opens with one question answered: what should I do next.
What to do next. Due decisions, applications ready to send, contacts to make, interviews coming up, and follow ups due, each as a row that opens the role it belongs to. Named contacts the scout retrieved this run land here too. Under that sits what the last scout run reported: how many sources it reached, how many roles it opened, how many cleared.
Five numbers. Active applications, meaning applied or further along. Your reply rate across everything you have sent, with a second figure for the applications you also wrote a human about. Median days to a first response. The average ghost risk sitting in your open pipeline. How many roles reached an interview or further. The app refuses to report a rate until there are enough applications behind it to mean anything, and says how many more it needs.
The pipeline is nine columns. Watching, Decided, Applied, Outreach sent, Screening, Interview, Final round, Offer, Closed. A role sits in exactly one of them, and inside a column the roles are ordered by how much they are worth pursuing rather than by when they arrived.
Every role is one record. A role enters once and carries everything after that: where it was found, the organization, the compensation and geography, the posting's requirement lines, the gates, the risk flags, the score and tier, the decision, which resume version went out, the contact, every message, the interview and the offer. Nothing is entered twice. Clicking a card opens a drawer with the whole record and its stage selector.
Possible duplicates get flagged, never deleted. Two postings for the same seat, one off a job board and one off the employer page, arrive with different links and both load. Two real openings at a large employer look identical here too. So the app names the pairs, shows you which source each came from, and leaves the removing to you.
Offer comparison and source performance. Once more than one role reaches an offer, the Board scores them side by side. Further down it reports which sources and which resume versions are actually producing replies, which is the number that should change where you spend next month.
Closing a role is permanent by design. A closed role stays on the board under Closed out with the reason on it, and nothing is deleted. It travels into the scout prompt as declined, so the scout stops resurfacing a requisition you already judged. It also cannot trigger a days elapsed action rule. The CSV export keeps the full history, closed rows included.
Board to CSV. One button, every role, every field, including the closed ones. Use it if you want the history in a spreadsheet or in a tracker you already keep.
Open Scout in the appOpens in a new tab so this guide stays where it is.
Scout does not search. It writes the instructions that make an AI assistant search properly, then hands you the file. The file is the product.
Two assistants, two outputs. Pick free ChatGPT and you get a prompt to paste into a new chat. Pick Claude and you get a plan file to download, which Claude Cowork can run on a schedule so a ranked shortlist is waiting in the morning. Everything else on the screen adapts to the choice. Paste the prompt, never upload it. An uploaded file gets read as a document to summarize and the search never runs, and the report that comes back looks correct and is empty of real work.
The gates are generated, not written by you. Your salary floor, your geography, your county list, your excluded organizations and industries, and your target level become the hard gates. Editing your profile updates them. There are ten, and a role that fails one is eliminated rather than scored down: evidence, confirmation, organization exclusion, geography, sector exclusion, role function, salary floor, salary scrutiny, duplicate, and eligibility.
Evidence and confirmation are the two that matter. Discovery is not evidence. A search result, a category page or a snippet finds a role and never proves one exists. Every reported role has to state one detail found only on that posting page, and scraped republishers never satisfy the gate on their own. Then confirmation: an aggregator page proves a listing existed, never that the seat is open, so every find gets checked against the employer's own careers page. Listed there means confirmed. A careers page that loads and does not list the role means the seat is closed and the role is eliminated. Unconfirmed is narrow and means one of exactly two things, a careers page that will not load or an organization with no careers page, and an unconfirmed role can never rise above Watch.
Coverage is counted, not described. The report opens with a run status line that states how many Priority 1 sources were reached out of how many, how many roles were opened, and how many cleared. Priority 1 is your job sites plus employer career pages as a single entry, and reaching fewer than two thirds of them marks the run Incomplete. An Incomplete run is forbidden from describing the market at all. Low coverage and a thin market produce identical empty reports. Only the count tells them apart.
Scoring and tiers. One hundred points across five factors: title and scope alignment, sector and mission fit, how well your proof points apply to the posted requirements, compensation fit, and conversion likelihood. Eighty five and above is Apply Today. Seventy to eighty four is Strong Consideration. Sixty five to sixty nine is Watch and Investigate. Below sixty five does not surface. Flagged for Review is a separate section for verification uncertainty and never for a gate failure. The payload block carries every scored role in every tier, so a Watch role still reaches your board with its verification questions still open.
The stop rule. Three consecutive runs with an empty Apply Today tier and the scout must state plainly that the posted market is exhausted and name exactly one change. New sources, a new lane such as direct outreach, or a revision to your criteria. One, not a menu. The count now travels inside the payload block, so it is no longer a number you have to keep by hand.
The self check runs on every report. Twenty items. It recounts the coverage arithmetic, audits every URL for a search pattern, checks that no excluded organization appeared anywhere, tests every role's score against its tier band, looks for the same seat reported twice under different sources, confirms every scored role reached the payload block, and reconciles the counts. It prints only the defects. A clean run prints one line.
Open Decide in the appOpens in a new tab so this guide stays where it is.
Decide answers one question per role: does this posting earn an hour of your tailoring time.
Paste the report and the board fills itself. Copy the whole scout report and paste it in. Every role, every action due, and every contact loads at once. The facts, the gate answers, the risk flags and the posting's own requirement lines arrive already in place, so you rate rather than copy and paste. You can also load a report file directly.
Pasting the same report twice loads nothing twice. Intake dedupes against the whole board, closed roles included, keyed by the posting URL with tracking parameters stripped and falling back to organization plus title when a URL is missing. Two different URLs at the same organization with the same title are two requisitions and both load, because they usually are.
Near misses come across on their own. Four of the scout's gates are matters of fact: a closed seat, an excluded employer, the wrong city, a posting you are not eligible for. Those stay out unless you ask for them with the checkbox, because reading them is not a decision, it is just work. Three are judgement calls with soft edges: salary, sector, and level. A role the scout eliminated on one of those loads anyway, lands in Watching marked as a near miss, and arrives with that one gate already answered No on the worksheet. A role three percent under your floor is worth your own eye.
The live gate is yours. The scout fills five of the six hard gates. It never fills the first one. Confirming a seat is open means opening the employer's careers page today and finding it listed, and no tool can do that on your behalf hours after the report was written. Until you answer it, the call cannot rise above Investigate First.
The evidence map. Each required line from the posting gets rated against your actual proof: Strong, Partial, Missing, or Unknown. Strong requires a named proof beside it, and if you cannot name the organization, the scale, the outcome and roughly the year, the line is not Strong. Unknown scores as zero on purpose, because an unverified claim is not evidence. The lines arrived copied off the page and the banner tells you the date the scout opened it. If a line is not in the posting, delete it. Anything you rate Strong with a named proof becomes the opening line of your outreach message in Connect.
Ghost job risk, out of a hundred. Eight signals, each weighted, plus points for an unconfirmed seat, a posting past sixty days, and an aggregator link you have not verified. The score and the count both feed the call.
One call comes out. Apply, Investigate First, or Pass, from the rules in section 9. Alongside it sit two numbers: fit, which is how well you match, and worth, which folds fit together with how live the seat looks and how reachable a human is. Fewer than three required lines rated and the call means nothing yet, and the app says so rather than producing a number.
Open Apply in the appOpens in a new tab so this guide stays where it is.
Apply is a pre flight check on the document, not a resume builder.
There is no score, because no score exists. Applicant tracking systems parse a file into fields and a recruiter searches inside those fields. Any tool that hands you 78 out of 100 invented that number, and the fixes it recommends aim at a threshold nobody is measuring. They also match strings and not meaning. Grant administration and grants management are the same work and two different strings.
Two things cost the interview. A document the scanner could not read, and language that does not match what the recruiter typed into the search box. Apply checks both.
Coverage, line by line. The requirement lines carry over from Decide the moment you open the station, so nothing gets pulled twice. Each one arrives on Auto, which is the overlap the tool computed between the posting's words and your resume text, and shows its arithmetic. Covered, Partial and Absent are yours to set when you disagree. Only Required lines are rated, so a line sitting on Preferred that this employer actually requires needs switching before the check will run.
Seven document questions, five of them fatal. Contact details in the header or footer, a table, two or more columns, a text box or sidebar, and skills drawn as a graphic are the five that alone can lose the record. Custom section names and an embedded photo are flagged but not fatal. Every one of these is layout rather than writing, and Print a clean copy in your profile fixes all of them in one press.
Paste your resume as plain text. That is the parse test, not a workaround. If a section disappears when you export to text, the scanner lost it too. The station reads that text for the faults that only show up in an export: no email or no phone number in the readable text, missing Experience, Education or Skills headings, fewer than two four digit years, headings with a space between every letter, long runs of spaces where tab stops collapsed, a single line hundreds of characters long with no sentence breaks, and unreadable characters from a custom bullet font.
The fix list travels. Apply names the missing terms line by line and hands you three blocks: the fix list, the term gap showing their phrasing against yours, and a rewrite packet. The packet is the whole request for an assistant, carrying the posting lines, your resume, the coverage call on each line, the document faults kept separate from the language work, and rules that forbid inventing an employer, a date, a number or an outcome. It ends with a self check the assistant has to run on its own draft.
The return trip, in order. Read the change log before the resume, because a full document hides its own edits and things go missing quietly. Paste the revised resume into your profile as a new version. Print a clean copy from it, since no rewrite closes a layout fault. Then come back, because the check reads whatever text your profile holds and the numbers move on their own.
When the gap will not close. A rewrite closes a Partial, because a partial is their wording for something you already did. It cannot close an Absent line, because absent means the work is not in your history and the packet forbids writing it in. The station stops asking for another packet once a revised resume has come back with a required line still absent. Three honest endings and no fourth: write the bullet yourself if you did the work and it simply is not on the page, apply with the gap and name it yourself in Connect before a scanner names it for you, or close the role out and put that hour on a seat you can win.
Open Connect in the appOpens in a new tab so this guide stays where it is.
Connect turns the hardest part of a search, writing to a stranger, into a decision about recipient and tone rather than a blank page.
Thirteen templates, and they are not interchangeable. Hiring manager after applying and hiring manager before applying. Someone you know there, either as a referral ask or for intel only. Recruiter or screener. Reconnecting with a dormant contact. An informational conversation request. A post interview thank you. Follow up one and follow up two. Withdrawing and closing the loop. A negotiation opener. A reply to a rejection. Then the relationship, the tone, the length, and whether it is an email, a LinkedIn message or a connection note.
The rules baked into every draft. Lead with a result that carries a number, not with a request. One line proving you actually looked at the organization. Ask for fifteen minutes and never for the job. Seventy five to a hundred fifty words. The proof points come straight from the lines you rated Strong in Decide, so a message with nothing concrete in it is telling you the evidence map is thin.
Copy and go. You get a message to copy and an appropriate next action, opening LinkedIn or an email. Nothing sends on your behalf. The app never looks up a person, never guesses an email address, and never touches a data broker. It gives you the search paths and writes the message once you name the recipient yourself.
Follow ups are tracked and capped. Log the send and both follow up dates set themselves: four business days, then seven days after that, then stop. Marking a reply ends the sequence and feeds the reply rate on your Board. The two follow up guardrail stands. Past that you are not persistent, you are a problem, and the answer is a different organization rather than a third message.
What stays human. Confirming you actually have a relationship with someone, and adding the personal hook. The app will not invent either, and a message that claims a connection you do not have is worse than no message.
Open Prepare in the appOpens in a new tab so this guide stays where it is.
Generic interview prep was built for software engineers. Prepare is built for the other room, and it runs from the first screen through the offer call.
The seat. Set the stage, from a first screen through a manager conversation, a panel, a board interview, or a final round. The sector lane is guessed from the posting text and you can override it. Change either one and the question set rebuilds.
The question set builds itself, and it starts with their own words. The posting's required lines come first, each one carrying what your resume actually does with it, so the likeliest question in the room is the one the station asks first. Then the core set, then the set for this stage, then the lane. Eighteen cards, because eighteen is the most anyone finishes in one sitting. Every card shows what is being tested underneath the question, which is the part that changes the answer.
Every answer takes four fields. Context, the number, what you decided, what changed. A card stays marked thin until the number is in it, because that is the field people skip and it is the field that decides the interview.
Questions to ask back. Grouped by what the answer reveals rather than by how clever they sound. Pick four or five, and add your own.
What to listen for. A checklist of warning signs to mark during the conversation or right after, never before. Three or more and you are looking at a seat that will not hold.
A prep sheet and a debrief. Print the sheet the night before. Fill the debrief within an hour of walking out, while you still remember what they asked and where you were weak: what they asked that you did not expect, where you were thin, what you promised to send, their stated next step and date, and your read on the seat. It lands on the role record with everything else.
The offer. A number is not an offer. Nine things are, and each gets rated from what you actually know, with anything unknown left at the middle mark rather than guessed high. Write your walk away line before you negotiate, not during. A number and a condition. Once two roles reach an offer, the Board scores them against each other.
No live coaching, by design. An earpiece or a second screen feeding answers is detectable, and in a sector this small it ends processes. All of the work happens before the room.
Decide, in the order the rules fire. The first one that matches is the answer.
| Condition | Call |
|---|---|
| Any hard gate answered No | Pass |
| Any of the six gates still unanswered, or fewer than three required lines rated | Incomplete, the call means nothing yet |
| Five or more of the eight risk signals present | Pass |
| Ghost risk 70 or above out of 100 | Pass |
| Two or more required lines rated Missing | Pass |
| Match rate on required lines below 60 percent | Pass |
| Any gate answered Unknown | Investigate First |
| Three or more risk signals, or ghost risk 45 or above | Investigate First |
| Any required line rated Unknown | Investigate First |
| Match rate below 80 percent | Investigate First |
| Everything above clear | Apply |
Match rate is weighted, not counted: Strong is worth one, Partial six tenths, Missing and Unknown zero.
Apply
| Condition | Call |
|---|---|
| No resume in your profile, or fewer than three required lines rated | Incomplete, the call means nothing yet |
| No email or no phone number in the readable text | Rewrite |
| Headings with a space between every letter | Rewrite |
| Any fatal document flag: contact in the header, a table, columns, a text box, or skills as a graphic | Rewrite |
| Three or more required lines Absent | Rewrite |
| One or more Absent, or three or more Partial | Fix first |
| Any nonfatal document flag still open | Fix first |
| Everything covered, no flags, contact details readable | Send as is |
One exception overrides the headline. Once a rewrite has already come back and a required line is still Absent, the station stops saying rewrite and says the call is yours, because telling somebody to do the thing they just did is how a tool loses their trust.
Prepare
| Condition | Call |
|---|---|
| Fewer than five complete cards | Not ready |
| Fewer than five cards carrying a number | Not ready |
| Under eight complete, or under six carrying a number | Thin |
| Fewer than three questions prepared for them | Thin |
| Eight or more complete, six or more with a number, three or more questions to ask | Ready |
One button on the top rail opens the list. Everything on it is built in this browser and nothing is uploaded.
| Item | What it is |
|---|---|
| Application packet | Every role still in play, with the decision and why, the posting, your proof against each requirement, contacts, interview cards and the timeline. Closed roles left out. |
| Board to CSV | Every role as a spreadsheet, closed ones included. |
| Career file backup | Everything this browser holds as one file. Restore it here or on another machine. |
| Decide, Apply, Connect and Prepare records | The printable record for the role you are working, one per station. |
| Scout prompt or plan file | The whole search brief, copied or downloaded. |
| Resume rewrite packet | The whole rewrite request for the role you are working. |
| Employer research prompt | A prompt that finds employers worth watching, built from your industries and your geography. |
| Clean copy of your resume | Your resume rebuilt in the structure a scanner reads. Not one word rewritten. |
Daily, unattended. The scout runs, reads your pipeline, scans the sources, and saves the report.
Daily, ten minutes. Open the Board and read what to do next. Paste the report into Decide. Work the roles that scored 70 and above, confirming each one is still live on the employer site and rating the requirement lines that arrived with it.
Same day. Apply on Apply calls. Log Passes so they never resurface. Put open items on the calendar.
Before every send. Run the posting and your resume text through Apply and clear the fix list. Five minutes, and it catches the failure that never produces a rejection email.
After every send. Open Connect and write to a human. An application with no message behind it is a lottery ticket.
The night before an interview. Print the prep sheet from Prepare. Fill the debrief within an hour of walking out.
Weekly. Back up your Career File. Read the five numbers on the Board and check whether the stop rule has fired.
Monthly. Revisit your profile and your sources. Read which sources and which resume version are producing replies. A stalled search usually needs a new lane, not more runs of the same scan.