job-feedlisted
Install: claude install-skill vikas9dev/claude-job-hunt
# Job Feed — Bulk Discovery + Triage
This skill is the cheap **Stage-1 triage gate** between bulk discovery and the expensive deep analysis in `/job-match`.
```
Stage 0 Pull JobSpy MCP or watchlist ATS feeds → 50-100 raw listings
Stage 1 Triage ← THIS SKILL — ZERO web searches:
role/YOE/location gates · skip-list · URL-first dedup · keyword pre-score
→ ranked shortlist (~5-15)
Park ⏳ Pipeline rows, JD text + posted_date kept
Stage 1b Drain ← THIS SKILL (`/job-feed pipeline`) → age-ranked, stale ones expired
Stage 2 Deep analysis /job-match on the top hits → full research + verdict
```
**Why the gate exists:** `/job-match` spends about 9 web searches per job. Running it on 100 raw listings means about 900 searches. Stage 1 is pure text work on JDs you already pulled. It costs almost nothing and protects the deep-analysis budget. Never pipe raw pull output straight into `/job-match`.
---
## Step 0 — Load the profile
Read **`.claude/skills/_shared/PROFILE_AND_FILTERS.md`** at the start of every run.
- If it does not exist, **stop**. Tell the user: `No profile found. Run /setup-profile, or copy PROFILE_AND_FILTERS.template.md to PROFILE_AND_FILTERS.md and fill it in.`
- If it still has `<<…>>` placeholders in a section this skill needs (§2, §4, §6, §7, §8, §10), list them and stop.
- Never copy values from the profile into this skill or into a sub-