opportunity-hunter

name	opportunity-hunter
description	Search for and rank job openings, grants, prizes, and fellowships relevant to the user's profile. Produce a structured markdown report saved to disk.

Opportunity Hunter

Search for and rank job openings, grants, prizes, and fellowships relevant to the user's profile. Produce a structured markdown report.

Trigger

Run when the user asks to find opportunities, scan the job market, or check for grants/fellowships/prizes.

Setup note: This skill reads the user's profile from ~/.claude/CLAUDE.md. Make sure your CLAUDE.md has a section describing your role, fields, target positions, career stage, and (if applicable) immigration/visa status. Replace {{REPORTS_DIR}} below with where you want reports saved.

---

Step 0: Load Profile

Pull candidate profile from ~/.claude/CLAUDE.md. You need at least:

• Name
• Current role
• Fields (research areas, disciplines)
• Methods (experimental, computational, theoretical, etc.)
• Target positions (academic TT? industry research? think tank? specify departments and seniority)
• Career stage (years post-degree, current rank)
• Immigration status (citizenship, visa, pathway) — if relevant. Cap-exempt employers, H-1B sponsorship requirements, roles requiring US citizenship/clearance all matter here.

If ~/.claude/CLAUDE.md lacks these details, ask the user once, then save to CLAUDE.md before continuing.

---

Step 1: Search (WebSearch — ~50 queries)

Run WebSearch queries across 5 categories. Use the current date to make queries timely (e.g., "2026-2027 academic job market"). Vary phrasing to maximize coverage. Substitute the user's fields and target positions into the query patterns below.

Category A — Academic Jobs (TT)

Search for tenure-track assistant professor positions in the user's target departments (e.g., marketing, management, OB, decision sciences, behavioral science, consumer behavior, or whatever fields they listed).

Target query patterns (substitute {FIELD} with each target field):

• "assistant professor" {FIELD} 2026 2027
• tenure track {FIELD} faculty job
• {FIELD} faculty position
• Check discipline-specific job boards (e.g., INFORMS, AMA, AOM, SJDM, Econ JOE, PhilJobs, etc.)
• Aggregators: HigherEdJobs; AcademicKeys; Chronicle

Category B — Industry Research

Search for research scientist / research engineer roles at labs that match the user's fields:

• Major AI labs (Anthropic, OpenAI, DeepMind, Meta FAIR, Microsoft Research, Apple ML)
• Tech research orgs (Spotify Research, Netflix Research, Amazon Science, Uber, Airbnb)
• Think tanks relevant to the user's area (RAND, Brookings, BIT, ideas42, NBER, etc.)

Query patterns:

• research scientist {FIELD} [company] 2026
• applied research {FIELD} tech
• {FIELD} researcher position

Category C — Grants

Search for grants matching the user's field and career stage. Major funders vary by discipline — examples:

• NSF directorates (SES, DRMS, DBI, etc.)
• NIH mechanisms (R21, R01, K-awards)
• Foundation grants (Russell Sage, Templeton, Spencer, Sloan, Kauffman, SSRC, etc.)
• Early-career fellowships (Sloan, NSF CAREER, K99/R00)
• Internal / institutional seed grants

Query patterns:

• {FIELD} grant 2026 deadline
• early career grant {FIELD}
• NSF CAREER {FIELD} deadline

Category D — Prizes & Awards

Search for:

• Early-career awards in the user's discipline
• Best-paper awards in target journals
• Rising-star awards from relevant societies
• Dissertation awards (if still eligible)

Query patterns:

• early career award {FIELD} 2026
• best paper award {FIELD} 2026
• rising star {FIELD}

Category E — Fellowships & Visiting Positions

• CASBS, Berkman Klein, Schmidt Futures, Radcliffe, Santa Fe Institute
• Visiting scholar programs at top departments
• Sabbatical-replacement visiting positions
• Discipline-specific fellowships

---

Step 2: Link Verification & Deep Fetch (WebFetch — ALL results)

CRITICAL: Verify every link before including it in the report.

For EVERY opportunity you plan to include, you MUST WebFetch the URL to confirm:

1. The page actually loads (not 404, not a generic careers page)
2. The specific listing/program described actually exists at that URL
3. The details (title, deadline, eligibility) match what the search snippet said

If a link fails or doesn't match:

• Try to find the correct URL via a follow-up WebSearch
• If you still can't get a working link, mark it as 🔗 Link unverified — check [organization website] and provide the org's main URL instead
• NEVER include a dead or unverified link without flagging it

For the most promising ~20 results, do a deeper fetch to extract:

• Full job/grant description
• Confirm deadline (flag if unverifiable)
• Check eligibility criteria (career stage, citizenship, field)
• Note immigration/visa information if relevant to the user
• Get application requirements (what materials are needed)

Common link pitfalls to watch for:

• Job board links expire quickly — always verify
• Company career pages use JS rendering that WebFetch can't parse — provide a search path instead (e.g., "Search careers.microsoft.com for '{FIELD}'")
• Grant program URLs change year-to-year
• Academic job board aggregators may link to expired postings

---

Step 3: Score Each Opportunity

Rate each opportunity 1-10 using this weighted rubric. Adjust weights if the user's CLAUDE.md specifies different priorities.

Criterion	Default weight	Description
Field match	25%	How well does the role/grant align with the user's research?
Career stage	20%	Is this appropriate for the user's current career stage?
Immigration compatibility	20%	Can the user apply given their visa/citizenship status? (Skip this criterion if not applicable; redistribute weight across others.)
Prestige / impact	15%	Institutional reputation, career signal value
Deadline feasibility	10%	Is there enough time to prepare a strong application?
Compensation / funding	10%	Salary, grant size, or award amount

Composite score = weighted sum, rounded to 1 decimal.

---

Step 4: Generate Report

Create a structured markdown report with these sections:

Report Structure

# Opportunity Report — [YYYY-MM-DD]

## Executive Summary
- Total opportunities found: X
- Top 3 highlights (with one-line reason each)
- Urgent deadlines in next 30 days
- Key seasonal note

## A. Academic Jobs (Tenure-Track)
| # | Position | Institution | Fit Score | Deadline | Link |
|---|----------|-------------|-----------|----------|------|

### Detailed Entries
For each:
- **Position:** [title]
- **Institution:** [name, department]
- **Link:** [URL]
- **Deadline:** [date — flag if unverified with ⚠️]
- **Fit score:** [X/10]
- **Why it fits:** [1-2 sentences]
- **Immigration notes:** [if relevant]
- **Application requirements:** [materials needed]
- **Recommended action:** [Apply / Watch / Skip — with reasoning]

## B. Industry Research
[Same table + detailed format]

## C. Grants
[Same table + detailed format]

## D. Prizes & Awards
[Same table + detailed format]

## E. Fellowships & Visiting Positions
[Same table + detailed format]

## Immigration-Flagged Opportunities (if applicable)
List all opportunities where immigration status is a concern or advantage, with specific notes.

## Upcoming Deadlines (Next 90 Days)
| Date | Opportunity | Category | Action Needed |
|------|-------------|----------|---------------|

## Methodology
- Date of search: [date]
- Number of queries: [N]
- Sources checked: [list]
- Limitations: [e.g., some deadlines may have changed]

---

Step 5: Save Report

Save the completed report to:

{{REPORTS_DIR}}/[YYYY-MM-DD]-opportunity-report.md

---

Step 6: Notify (optional)

If the user has the notify-me skill installed, trigger it with a short summary:

Opportunity report ready: [X] opportunities found. Top pick: [best one]. [N] deadlines in next 30 days.

---

Anti-Hallucination Rules

These are critical. Violating them makes the report useless.

1. Never fabricate a listing. Every opportunity must come from an actual WebSearch or WebFetch result.
2. Flag unverified deadlines with ⚠️. If a deadline cannot be confirmed from the source page, say so explicitly.
3. Do not guess application URLs. Only include links you actually retrieved AND verified via WebFetch.
4. Every link in the final report must be WebFetch-verified. If you could not load the page, mark it with 🔗 Link unverified and explain how to find it manually.
5. If a search returns no relevant results for a category, say so. Do not fill the section with tangentially related results to look comprehensive.
6. Distinguish between confirmed open positions and recurring annual opportunities where the current cycle may not be open yet.
7. Never construct URLs by pattern-matching.
8. If WebFetch fails on a JS-heavy page (common for Ashby, Greenhouse, Workday, Lever), note this and provide a manual search path.

---

Seasonal Awareness

The academic job market is highly seasonal (at least in the US):

• Aug–Dec: Peak TT job postings (most schools post Sept–Nov)
• Jan–Mar: Flyouts and offers
• Apr–Jun: Late-cycle and failed-search re-postings
• Year-round: Industry roles, some grants

Interpret search results accordingly. If running outside peak season, note which recurring opportunities to watch for and when they typically open.

---

Design Principles

• Err on comprehensiveness. Include borderline-relevant opportunities. The user can cut; they can't find what you didn't show them.
• Immigration is a dealbreaker if flagged. If the user has immigration constraints in CLAUDE.md, respect them rigorously.
• Actionability over description. Every entry should end with a clear recommended action.
• Real links only. Every URL must be from an actual search/fetch result.