filter-jobs
Skill1 starsChanged 3 months agoMITView on GitHub
Score and tier the current job queue against the user's CV and profile, flag language and work-authorisation risk, and write the results to the tracker. Runs automatically at the end of find-jobs; use it on its own to re-tier an existing queue after changing the CV, the profile, or the filter rules.
In this repository4 other files it gives its agents.
- apply-to-jobSkill
- find-jobsSkill
- follow-upsSkill
- setupSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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 77
If the workbook is open in Excel it will be locked.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Filter jobs against this CV
The pipeline matched title words and location strings. That is all it can do. Everything that needs reading comprehension happens here.
find-jobs calls this automatically — you rarely need to run it by hand. Do run it on its
own after the user edits their CV, changes profile.md, or wants a queue re-judged under
new rules.
Read first
profile.md— work authorisation and language level. These decide eligibility, and getting them wrong costs the user a week.cv/master_cv.texor.html— their actual experience. You are scoring against this, not against a job family label.data/queue_<today>.json— the roles to judge. If today's file is missing, take the most recent.Job_Tracker.xlsx— what is already there, so you do not add it twice.
Tier each role
A — apply now. Fits their experience, no eligibility problem, and they could plausibly get an interview. Be selective. If everything is tier A, the tier means nothing.
B — a real stretch. Above their seniority, an adjacent function, or a domain they would have to argue for. Worth it when A runs dry.
C — probably blocked, check first. A language bar, an on-site requirement they cannot meet, a clearance. Do not spend an application on a guess — but do not silently bin it either, because the guess is sometimes wrong.
X — they cannot take it. Wrong country for their work authorisation, or a stated citizenship requirement. Say which.
Two things the pipeline cannot see
Language risk. Infer it from the posting title, which is objective and free: markers
like (H/F), (F/H), (M/W/D), CDI, CDD mean it was written for a local market and
the role probably runs in that language. Record the evidence, not just the verdict — "F/H
in the title" is checkable, "seems French" is not.
Work authorisation. Compare the role's country against what profile.md allows. A
"remote — Europe" posting still usually needs the right to work in one specific country;
treat that as unresolved rather than fine.
Where you disagree with the score, say so
A low-scoring role at a company whose domain matches the user's background is often worth more than a high-scoring generic one. The pipeline has no way to know that a hospitality degree makes a hotel-software company interesting. You do. Put it in the notes and raise it in your summary.
Equally: a role that scores well but sits three levels above them is a B at best. Score is an input, not a verdict.
Write the tracker
Append to the Applications sheet in Job_Tracker.xlsx with openpyxl. Preserve existing
rows and any Status the user has already filled in — never rewrite the sheet wholesale.
# | Tier | Company | Role | Location | Work auth | Score | Source | Job link | Status |
Date applied | CV file | Notes
- Dedupe against what is already there by company and role before appending, and report how many were new versus already known
Statusstarts atNot started- Colour the tier cell: A
C6EFCE, BFFEB9C, CFCE4D6, XE7E6E6 - Pair every row with its own link before you sort. Sorting rows and zipping them
against an unsorted list of URLs attaches every link to the wrong job. Build
(row, url)pairs first, then sort the pairs. Check two rows by hand afterwards.
If the workbook is open in Excel it will be locked. Say so and ask them to close it — do not write somewhere else instead.
Report
Tier A in full, notable Bs, counts for C and X, and anything where you disagreed with the score. Say how many were new. Then point at the best one:
/apply-to-job <url>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.

