triage-jobs
Skill7 starsChanged 3 days agoMITView on GitHub
Triage the new jobs in Openings by the owner's Preferences and career entries - "go through new jobs", "triage my new jobs", "filter them well", "anything good today?". Shortlists clear fits, dismisses confirmed mismatches with a reason, resolves duplicates and asks about the rest.
In this repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- maintain-careerSkill
- manage-sourcesSkill
- openingsSkill
- prepare-applicationsSkill
- prepare-interviewSkill
- setup-openingsSkill
- track-applicationsSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 11
Read openings first. Triage only: shortlisting starts an application, but documents are prepare-applications work.
- Very specific: line 17
Call workspace_context. Before reading any job, write down for yourself:
- Very specific: line 40
Use the jobs_list filters to find groups that likely fail a dealbreaker: a required language the owner does not speak, postings no longer …
- Distinctive: line 48
Read every remaining posting in full, 20 at a time, and decide each:
- Distinctive: line 61
Never decide from the title alone, and never dismiss to empty the list.
≈ “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
Go through new jobs
Read openings first. Triage only: shortlisting starts an application, but documents are prepare-applications work.
1. Know the criteria
Call workspace_context. Before reading any job, write down for yourself:
- Dealbreakers from What I rule out and Facts about me (work permit, places, languages, seniority, domains);
- Must-haves and pluses from What I'm looking for, and what the career entries show the owner has done;
- Lessons from
feedback: a restored dismissal means that reason was too strict, a closed application from your shortlist means that kind of job is not wanted; - Rules from Working with me, such as asking before large dismissals.
Missing evidence of a skill is not proof the owner lacks it. Read a career
entry in full (career_get) when a decision turns on it.
2. Take the cohort
List the new jobs with jobs_list, with counts, and page through all of them
before writing, so your cohort does not move while you work. Note how many
there are and what the signals say (required languages, no longer listed,
likely duplicates).
3. Clear whole kinds of mismatch
Use the jobs_list filters to find groups that likely fail a dealbreaker: a
required language the owner does not speak, postings no longer listed, a ruled
out title or place. A signal is a lead, never a verdict: read each posting
(jobs_get, 20 at a time) and dismiss only when the text confirms it. A
posting written in another language does not by itself require it.
4. Read the rest
Read every remaining posting in full, 20 at a time, and decide each:
- Shortlist: it meets the must-haves, breaks no dealbreaker, and the owner's record makes a credible case. Be strict: the owner will get a CV and letter for each one.
- Dismiss: a concrete dealbreaker or a clear mismatch with what they want, stated in the posting. The reason is one sentence the owner reads beside the job: the fact and their preference, as in "Requires fluent German; you speak English and Italian" or "Sales role; you rule out non-engineering jobs".
- Leave it new and ask: a plausible fit that depends on a fact you do not have, or a borderline case. Group these into a few questions.
Never decide from the title alone, and never dismiss to empty the list. A posting without its full text cannot be judged: leave it and say so.
5. Write in batches
Send the decisions with jobs_triage, up to 200 at a time, and check every
item; a failed one stays where it was, so fix the cause or mention it. When
Working with me sets a limit, show the owner a sample first.
6. Duplicates
For each job you shortlist, and each likely duplicate, check the vacancy:
- The same vacancy: same employer and requisition or apply link, or the same title, place and duties in nearly the same text.
- Different: another requisition, level, city or team, or a reopened round months later. Same company and title alone prove nothing.
With jobs_link (decision same), the copy is job_id and the record to keep,
the one with the application or history, is other_id: other_id survives.
Or mark the pair different so the flag stops. When unsure, leave both and ask.
7. Check the result
List the new jobs again with counts. Say none are left only when that read shows it, and mention jobs that arrived meanwhile.
Report back
In chat: shortlisted, dismissed and left counts; the shortlisted jobs by title and company with one line of fit each; the most common dismissal reasons; your questions for the jobs left; anything a source keeps sending that manage-sources could filter. Offer to prepare the new shortlist.
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.

