evaluate

Evaluate — score one job ad

Scores a single job ad against the user's CV and ranked preferences and gives a clear apply / don't verdict. For many ads at once, use jobhunt:search instead.

Inputs

One job ad — pasted text, a URL, or a screenshot.

Reads / writes

• Reads: ~/.swissjobs/cv.md, ~/.swissjobs/preferences.md. If either is missing, tell the user to run jobhunt:setup first.
• Writes: optionally append the posting to ~/.swissjobs/seen-postings.jsonl.

Steps

1. Load the CV and preferences (especially the three ranked job types).
2. Obtain the ad text. If given only a URL, read it via web fetch, or open it in Claude for Chrome (the user's own session) if it needs authentication. Respect site terms — see ${CLAUDE_PLUGIN_ROOT}/shared/references/boards.md.
3. Detect the ad's language (DE/FR/IT/EN) and understand it natively. Write your output in the user's primary language (default EN) unless asked otherwise.
4. Map the ad to the best-matching ranked job type (1, 2, 3, or "none").
5. Requirement-by-requirement: list each key requirement and mark Met / Partial / Gap, citing concrete CV evidence. Judge only against what is truly in the CV — no fabrication (${CLAUDE_PLUGIN_ROOT}/shared/references/guardrails.md).
6. Score the fit 0–100 with brief reasoning. (TODO(you): adjust the "worth applying" threshold to taste; default ~65+.)
7. Positioning: 2–4 truthful angles for how the user should present themselves for this role.
8. Verdict: a clear "worth applying or not", in priority context (e.g. "Matches priority 2; strong fit — apply").
9. Offer next steps: jobhunt:tailor-cv, jobhunt:draft-email, jobhunt:interview-prep; offer to log the posting to the seen list.

Output format

• Header: role · company · location · ad language · matched priority
• Fit score (0–100)
• Requirements table: requirement | Met / Partial / Gap | evidence
• Positioning bullets
• Verdict + recommended next step