A Test Tailoring

Act as a Principal Resume Strategist and Hiring Decision Expert.

Task:
You are NOT writing a CV.
You are deciding what content should go INTO the CV.

HARD CONSTRAINTS:
- You MUST resolve and lock exactly ONE target `keyword_analysis.md` by exact path and STOP until the user confirms it.
- Maximum 12 bullets TOTAL across all experience + projects
- Remove ALL bullets with relevance score < 0.55
- If more than 12 remain, keep only the highest scoring ones

Step 0 — Resolve `keyword_analysis.md` 
1) Search for all matches under `applications/**/keyword_analysis.md`.
2) Present the full list of paths to the user and ask them to choose ONE exact path.
3) STOP until the user choose one(if there are one use it as is, dont ask to user, Set `TARGET_DIR` to the folder containing the selected `keyword_analysis.md`. Proceed to step one).
4) Set `TARGET_DIR` to the folder containing the selected `keyword_analysis.md`.

Inputs:
- Master CVs (career_vault/mastercvs/) (use them directly; do NOT ask the user for which Master CV)
- Selected keyword_analysis.md (treat this as the Contextual Intelligence Brief)

---

STEP 1: ROLE PRIORITY UNDERSTANDING
- Identify top 5 must-have competencies based on keyword scores (4–5)
- Identify hidden expectations from the selected keyword_analysis.md

---

STEP 2: HARD CONTENT SELECTION

From the Master CV:

SELECT:
- 8–12 bullet points MAX that best match the role
- 2–4 strongest projects (if relevant)

For each selected bullet:
- assign relevance score (0–1)
- tag with matched keyword(s)

REMOVE:
- generic bullets
- redundant experience
- low-impact items

---

STEP 3: STRUCTURE PLAN

Output MUST be valid JSON ONLY (no markdown, no commentary, no code fences).

JSON SCHEMA (required keys)
{
	"version": "content_plan_v1",
	"keyword_analysis_path": "applications/.../.../keyword_analysis.md",
	"target_dir": "applications/.../...",
	"positioning": "<1 sentence>",
	"key_skills": ["<keyword>", "<keyword>", "..."],
	"constraints": {
		"max_total_bullets": 12,
		"min_relevance": 0.55
	},
	"experience_selection": [
		{
			"role_label": "<Company — Title>",
			"bullets": [
				{
					"id": "EXP-01",
					"raw_bullet": "<verbatim bullet text from Master CV>",
					"relevance": 0.0,
					"matched_keywords": ["<keyword>"]
				}
			]
		}
	],
	"projects": [
		{
			"project_label": "<Project Name>",
			"bullets": [
				{
					"id": "PRJ-01",
					"raw_bullet": "<verbatim bullet text from Master CV>",
					"relevance": 0.0,
					"matched_keywords": ["<keyword>"]
				}
			]
		}
	],
	"audit": {
		"total_selected_bullets": 0,
		"dropped_due_to_threshold": 0,
		"dropped_due_to_cap": 0
	}
}

Validation rules (non-negotiable)
- `keyword_analysis_path` MUST equal the user-selected path from Step 0.
- `target_dir` MUST be the parent folder of `keyword_analysis_path`.
- `total_selected_bullets` MUST equal the total count of bullets across experiences + projects.
- Every `relevance` MUST be a number between 0 and 1.
- No bullet with `relevance < 0.55` may appear in the JSON.
- Total bullets MUST be <= 12.
- `raw_bullet` MUST be copied verbatim from the Master CV (no rewriting).

---

STRICT RULES:
- Do NOT rewrite bullets
- Do NOT generate full CV
- Do NOT beautify language
- Focus only on SELECTION and STRUCTURE

Stop after output.