llm-job-pipeline: Command for Claude Code

/jobs-profile

Manage config/user_profile.md — the single source of truth for scoring AND filtering. Three modes selected by the first argument:

Invocation	What it does
/jobs-profile (no arg)	Print a profile summary + active HARD filters
/jobs-profile rules	Edit geo policy (ban_regions, keep_countries, ban_us_only, …) + exclude_title_keywords in ## HARD_FILTERS
/jobs-profile edit	Broader edits — seniority, target roles, exclude patterns, geography

Profile creation (first-time) happens on the onboarding page (docs/index.html) or automatically when you first run /jobs-new. This command never creates the profile from scratch — it only reads or edits an existing one.

---

Mode: view (no arg)

Step 1 — Locate the profile

Check whether config/user_profile.md exists.

• File present — proceed to Step 2.

• File absent — print:

  No profile found — showing example config/user_profile.md does not exist yet. The output below comes from config/user_profile.example.md so you can see what a completed profile looks like. To create your real profile, complete the onboarding wizard on the landing page (docs/index.html) or run /jobs-new (first run detects an empty database and walks you through setup).

  Then read config/user_profile.example.md and proceed — but treat every displayed value as illustrative, not real.

Step 2 — Print the profile summary

Read the profile and summarise in plain language:

• Seniority target (from ## SENIORITY or equivalent section)
• Target roles / titles (from ## TARGET_ROLES)
• Geography preferences (from ## GEOGRAPHY)
• Exclude patterns (from ## EXCLUDE_PATTERNS — soft score penalty)
• Hard filters — run the snippet below and read the result back conversationally:

python3 -c "
import sys; sys.path.insert(0, 'scripts')
from hard_filters import load_hard_filters
hf = load_hard_filters()
fmt = lambda v: ', '.join(v) if v else '(none)'
print('Banned regions:', fmt(hf['ban_regions']))
print('Kept (whitelist):', fmt(hf['keep_countries']))
print('Banned countries:', fmt(hf['ban_countries'] + hf['exclude_countries']))
print('Drop US/Canada-only roles:', 'yes' if hf['ban_us_only'] else 'no')
print('On-site no-penalty regions:', fmt(hf['onsite_ok_regions']), '| penalty:', hf['onsite_penalty'])
print('Title words dropped:', fmt(hf['exclude_title_keywords']))
"