job-triage
Skill1 starsChanged 2 months agoMITView on GitHub
Turn a raw list of job postings into a short, honest short-list by screening the requirements block for hard gates before anything reaches the digest. Use when triaging scraped or pasted postings, refreshing a pipeline, or producing a daily job digest. Screens on stated requirements only, never on responsibilities.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- Ben-Bot-22/job-hunt-kittheir own version, first seen Oct 9, 2026
- AkbarDevop/ai-job-agenttheir own version, first seen Apr 26, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- AkbarDevop/ai-job-agent★ 61same name
- Ben-Bot-22/job-hunt-kit★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- application-packageSkill
- apply-prefillSkill
- councilSkill
- cv-tailorSkill
- github-publishSkill
- heavy-build-protocolSkill
- human-voiceSkill
- jd-analyserSkill
- security-vetSkill
- wiki-syncSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
If results look suspiciously global, your location filter has broken.
- Distinctive: line 61
This does not touch the stated-language rule: Never infer a language bar from the fact that a posting is written in a given language — that …
- Very specific: line 66
Use this screen only where the mission block is unambiguous about the core deliverable.
- Distinctive: line 103
Keep it scannable. Recommend at most two or three for full evaluation.
- Distinctive: line 112
Never screen out on a soft preference: Only the gates above kill a role.
- Distinctive: line 116
Do not assume a scheduled scan ran: Check the newest timestamp in your source before claiming the data is fresh.
≈ “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
Job Triage
Raw postings in, a short-list of two or three worth real effort out — plus an auditable record of what got screened out and why.
This skill does not scrape. Bring it postings from wherever you get them: an export, a board, a pasted list, an ATS API. It does the judging.
Load first
profile.yml— target archetypes, exclusions, location rule, granted exceptionsapplications.tsv— so nothing already tracked resurfaces- A dedup history file, if you keep one
Steps
1. Coarse filter
Drop, before reading anything closely:
- Seniority mismatches. A search for apprentice roles should never surface "Senior Staff X"
- Excluded functions from
profile.yml - Locations outside the commutable zone, unless
profile.ymlgrants that role an exception - Anything already in
applications.tsvat any status
If results look suspiciously global, your location filter has broken. Check it before blaming the source.
2. The requirements-block screen
For each survivor, fetch the posting and read the requirements block only: "Requirements", "Your profile", "Qualifications", "Profil recherché", "Type d'études", "À propos de toi".
Apply jd-analyser's hard gates:
| Gate | Kill signal in the requirements block |
| Degree field | Names a field the candidate does not hold |
| Specialised school | Names a school type they are not in, as an exclusive |
| Stated language | "fluent French", "proficiency in X", "bilingual" above their level |
| Hard tool bar | A named tool listed as required with no evidence in facts.md |
| Portfolio requirement | A required portfolio, book, or prior placement in a craft the candidate has not done |
| Location | Outside the zone with no exception granted |
The mission-block screen
Two things are not gates, because nothing is stated, and still make a role unbuildable. For these, read the missions — and read them only as a description of the day job, never as an inferred requirement.
| Screen | Kill signal in the MISSION block | Why |
| Authorship as the deliverable | "write articles, newsletters, white papers", "editorial content", "SEO copy", "press relations", "editorial calendar" | The job is producing polished copy in a language the candidate does not write at that level. No CV craft closes that |
| Design tooling as the deliverable | "infographics", "banners", "high-fidelity mockups", "design system", "visual identity", "video editing", "brand assets" | Check facts.md honestly. Product-side UX is not graphic design, and a tool listed as an aspiration is not a skill |
This does not touch the stated-language rule. Never infer a language bar from the fact that a posting is written in a given language — that rule stands. This screen reads what the work produces, which is a different question from what the employer requires. One posting stated no language bar at all and was still unbuildable, because the deliverable was editorial copy.
Use this screen only where the mission block is unambiguous about the core deliverable. It is
deliberately narrow: two categories, both tied to something facts.md explicitly cannot claim.
A gate or screen hit means the role does not enter the digest. Log it under a Screened out section
with the one-line reason, so every decision the filter made stays auditable and the user can
overrule any of them.
Read the requirements block, never the responsibilities. Three roles in a row once had responsibilities matching the candidate well and were each disqualified by a gate. Responsibilities describe the job. Requirements decide whether you can hold it.
Only the stated bar counts. A posting written in French that names no French requirement does not get a French gate. Never infer a language bar from the posting's language or the company's country.
3. Rank the survivors
By archetype fit from profile.yml. Note contract length where the posting states it.
4. Write the digest
# Today — YYYY-MM-DD
N new roles found. M survived screening.
## Worth evaluating
1. **Company** — Role
url · source · duration · why it ranks
## Also live
...
## Screened out
- **Company** — Role — gate: stated language ("maîtrise du français")Keep it scannable. Recommend at most two or three for full evaluation. Everything else stays in the pipeline.
Rules that exist because of real mistakes
- Never present raw output as a short-list. Unfiltered feeds surface senior roles for a junior search. Filter seniority before showing anything.
- Never gate on contract duration. A shorter-than-preferred contract is a flag and a demotion. It is the user's decision, not the filter's.
- Never screen out on a soft preference. Only the gates above kill a role. "Ideally", "a plus", "nice to have" are BONUS and stay in the digest.
- Dedup against the tracker, not just the scan history. A role already at Applied must never reappear.
- Do not assume a scheduled scan ran. Check the newest timestamp in your source before claiming the data is fresh. A silently broken scheduled job once made a stale pipeline look current for weeks.
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.

