map-career-path

name	map-career-path
description	Map a transition from a current role to a target role — surface real LinkedIn-style transition examples + a personalized gap roadmap (skills, experiences, signals) grounded in the user's profile. Optional Google CSE script (better recall) when keys are present; WebSearch fallback ships keyless.
when_to_use	Use when the user is contemplating a role change ("PM → Director of Product", "SWE → AI Engineer", "Analyst → PM"). Lower priority than the apply / recruiter loops.
user-invocable	true
allowed-tools	Read, Write, Bash, WebSearch

map-career-path

Show the user how others made the move, and where they'd need to grow. Hard rules: ../_shared/RULES.md.

Inputs

1. current_role (required).
2. target_role (required).
3. Profile — profile/master-profile.md (for the gap analysis).
4. Optional region — --region US | IN (else use profile's target.regions).

Method

0. Resolve the search backend.

   • If GOOGLE_API_KEY + CSE_ID env vars are present →
     python3 scripts/career_paths.py --current "<X>" --target "<Y>" --region <code> (CSE-driven; better recall + structured snippets).
   • Else → WebSearch with the same 4 query templates the script uses.

   Either path returns up to 10 transition examples.

1. Query templates (the script + the WebSearch fallback use identical queries):

   • "<current>" "<target>" site:linkedin.com/in
   • from "<current>" to "<target>" site:linkedin.com
   • "<target>" "previously <current>" site:linkedin.com
   • career transition "<current>" "<target>" (broader, no site restriction)

2. Filter for relevance. Keep entries where the snippet/title shows both the current and target role (or near-synonyms — "Product Manager" matches "Sr. PM"). Dedup by canonical URL. Drop entries with no transition signal.

3. Synthesize patterns. Across the kept profiles, surface 3–5 common patterns (e.g., "most PM → Director moves went through a 1–2 yr 'Group PM' stint"; "AI Engineer transitions usually start with shipping one production LLM project at the current role"). Patterns must be observable in the snippets — don't generalize beyond evidence.

4. Personalize the roadmap against the user's profile:

   • You already have: items the profile evidences.
   • Build next: specific, measurable skills/experiences (e.g., "ship one LLM-eval doc in the current role"; not "get more AI experience").
   • Soft signals (optional): writing, speaking, OSS, certifications — only real, named ones.

Output

paths/<current-slug>__to__<target-slug>.md

# <Current> → <Target>
> as of <date> · source: cse | websearch · region: <code>

## Real transitions (up to 10)
- Name (or anonymized) — current company → target company — <link>
  - snippet quote (≤ 20 words)
- ...

## Common patterns
- ...
- ...

## Your roadmap

### You already have
- ...  (cites profile facts)

### Build next
- specific, measurable item
- ...

### Soft signals (if you want them)
- real, named resource
- ...

Stdout

3-line summary — # transitions found · 1 dominant pattern · 1 highest-leverage next step. Suggest /find-jobs filtered to bridge roles, then /score-fit.

Anti-patterns

• Treating a single snippet as authoritative ("most people do X" with N=1).
• Inventing roles, company names, or transitions not in the snippets.
• Generic "get an MBA / certification" suggestions without profile context.
• Quoting context.career_goal verbatim (paraphrased direction is fine, verbatim is not).
• Sourcing transition examples from anything other than LinkedIn-public snippets.
• Emoji.

Rubric

• ≥ 3 transition examples (or honestly report the small N when search returned less).
• Every transition has a link.
• Roadmap items are specific (no "get more experience").
• "You already have" cites actual profile facts.
• Graceful degrade without CSE keys (skill still produces a roadmap; just smaller N).

Next steps

Roadmap ready. Next: /find-jobs filtered to bridge roles, then /score-fit.