score-cv

name	score-cv
description	Score a tailored CV against a previously saved job analysis artifact. Use when the user wants a grounded 0-100 match score without re-running job extraction.
disable-model-invocation	true

/score-cv

Score a tailored CV against a saved job analysis artifact.

Usage

/score-cv --analysis <path-or-id> --cv <tailored-yaml-path>

Examples:

• /score-cv --analysis a7c19f2d --cv templates/david_alecrim_tempo_rust_engineer_cv.yaml
• /score-cv --analysis output/job_analyses/a7c19f2d__tempo__rust_engineer.json --cv templates/david_alecrim_tempo_rust_engineer_cv.yaml

Workflow

Follow these steps exactly in order:

Step 1 — Resolve the saved analysis artifact

--analysis <path-or-id> is required.

Resolve it from the locally saved analysis artifacts under output/job_analyses/.

If resolution fails, report the error and stop.

Step 2 — Resolve the tailored CV

--cv <tailored-yaml-path> is required.

If the file does not exist, report the error and stop.

Step 3 — Extract CV evidence

Read the tailored YAML and extract the relevant resume evidence directly in the workflow context.

If evidence extraction fails, report the error clearly and stop.

Step 4 — Score the CV

Spawn the score-cv-match agent. Pass it a JSON object with:

• job_analysis
• cv_evidence
• scoring_rubric

Use this exact rubric:

{
  "required_skills": 40,
  "responsibilities": 25,
  "ats_keywords": 15,
  "positioning": 10,
  "preferred_skills": 10
}

The scoring agent must return grounded JSON only.

Step 5 — Report

Return:

• final score
• category breakdown
• key matched evidence
• missing requirements

Non-negotiable rules

• Never re-run analyze-job-posting when --analysis is provided.
• Never score against hidden thread context.
• Only score from the saved analysis artifact plus extracted CV evidence.