jobhunt

Files (52)

• plugin.json · .claude-plugin/
• .python-version
• SKILL.md
• candidates-table.tsx · dashboard/components/
• embedding-map.tsx · dashboard/components/
• learning-plan.tsx · dashboard/components/
• opportunities-board.tsx · dashboard/components/
• opportunity-list.tsx · dashboard/components/
• pipeline-board.tsx · dashboard/components/
• schema-inspector.tsx · dashboard/components/
• skills-matrix.tsx · dashboard/components/
• stats-overview.tsx · dashboard/components/
• lib.ts · dashboard/
• page.tsx · dashboard/pages/jobhunt/collection/[id]/
• page.tsx · dashboard/pages/jobhunt/opportunity/[id]/
• page.tsx · dashboard/pages/jobhunt/
• page.tsx · dashboard/pages/jobhunt/position/[id]/
• layout.tsx · dashboard/pages/
• route.ts · dashboard/routes/candidates/action/
• route.ts · dashboard/routes/candidates/
• route.ts · dashboard/routes/collection/[id]/
• route.ts · dashboard/routes/embedding-map/
• route.ts · dashboard/routes/fit/
• route.ts · dashboard/routes/gaps/
• route.ts · dashboard/routes/learning/
• route.ts · dashboard/routes/notes/[id]/
• route.ts · dashboard/routes/opportunities/
• route.ts · dashboard/routes/opportunity/[id]/
• route.ts · dashboard/routes/opportunity/update/
• route.ts · dashboard/routes/pipeline/
• route.ts · dashboard/routes/position/[id]/
• route.ts · dashboard/routes/seekers/
• route.ts · dashboard/routes/skills/
• route.ts · dashboard/routes/status/
• validate_views.py · dashboard/
• collection.yaml · dashboard/views/
• company.yaml · dashboard/views/
• contact.yaml · dashboard/views/
• hub.yaml · dashboard/views/
• opportunity.yaml · dashboard/views/
• position.yaml · dashboard/views/
• embedding_map.py
• hooks.json · hooks/
• job_forager.py
• jobhunt.py
• pyproject.toml
• quality-checks.yaml
• schema-diagram.svg
• schema.tql
• skill.yaml
• SKILL.md · skills/jobhunt/
• uv.lock

SKILL.md (first lines)

# Job Hunting Notebook Skill

Use this skill to manage a job search as a knowledge graph. Claude acts as the career agent — browsing job boards, evaluating postings against the seeker's profile, and building structured understanding of positions, companies, and fit over time.

**Key principle:** The agent does the sensemaking. Scripts are tools that search APIs, store data, and run queries. The agent decides what's relevant, writes the rationale, and manages the pipeline.

**Role model:** The job seeker is a BFO/UFO *role* that inheres in a person. One person can have multiple roles (job seeker, author, evaluator). The role holds search-specific data; the person holds identity data.

---

## 1. Quick Start

### Prerequisites

- TypeDB must be running: `make db-start`
- Dependencies installed: `uv sync --all-extras` (from project root)
- Schema loaded: run `make build-db` after `make build-skills`

> **Path note:** Replace `.claude/skills/jobhunt` below with your installation directory:
> - **Claude plugin install:** `${CLAUDE_PLUGIN_ROOT}/skills/jobhunt` (self-contained bundle at `plugins/jobhunt/`)
> - **skillful-alhazen project:** `.claude/skills/jobhunt`
>
> When installed as a plugin, TypeDB starts automatically on session start (no manual init required).

### Environment Variables

- `TYPEDB_HOST`: TypeDB server (default: localhost)
- `TYPEDB_PORT`: TypeDB port (default: 1729)
- `TYPEDB_DATABASE`: Database name (default: alhazen_notebook)

### Essential Commands

```bash
# Ingest a job posting from URL
uv run python .claude/skills/jobhunt/jobhunt.py ingest-job \
    --url "https://boards.greenhouse.io/anthropic/jobs/123456" \
    --priority high

# List your pipeline
uv run python .claude/skills/jobhunt/jobhunt.py list-pipeline

# Show position details
uv run python .claude/skills/jobhunt/jobhunt.py show-position --id "position-abc123"

# Add a note to any entity
uv run python .claude/skills/jobhunt/jobhunt.py add-note \
    --about "position-abc123" --type research \
    --content "Strong AI safety focus, good culture fit"
```

### Command Output Pattern

`uv run` emits a `VIRTUAL_ENV` warning to stderr. Always use `2>/dev/null` when piping output to a JSON parser -- never `2>&1`, which merges the warning into stdout and breaks JSON parsing.

---

## 2. Job Seeker Profile

The job seeker is modeled as a **role** (BFO/UFO pattern) that inheres in a person. The role holds search-specific data (target role, industries, salary expectations) while the person entity holds identity data (name, email, linkedin).

### Create a Seeker Profile

```bash
uv run python .claude/skills/jobhunt/jobhunt.py create-seeker-profile \
    --person op-f25ab4b15b0f \
    --target-role "Principal AI Architect" \
    --industries "AI, Biotech, Scientific Computing" \
    --salary "200k-250k" \
    --location "Remote / SF Bay Area" \
    --focus "TypeDB-powered knowledge systems, AI for science"
```

This creates a `jhunt-job-seeker-role` entity linked to the person via `alh-role-bearing`. All pipeline data (opportunities, skills, candidates) scopes to this role.

### Build the Skill Profile

Skills are owned by the seeker role, derived from real evidence, and use a simplified Bloom's taxonomy for proficiency levels.

**Proficiency levels:**
| Level | Meaning | Evidence type | Interview readiness |
|-------|---------|---------------|-------------------|
| `expert` | Can design/teach | Publications, architected systems, led teams | Can whiteboard, critique alternatives |
| `practiced` | Hands-on experience | Projects, production use, coursework | Can walk through how you used it |
| `aware` | Conceptual knowledge | Read papers/docs, attended talks | Can answer "what is X" questions |
| `none` | Not known | No evidence | Cannot discuss |

#### Step 1: Ingest Profile Artifacts

The agent ingests the seeker's professional artifacts as sources of evidence. Each artifact is stored and linked to the seeker via `alh-aboutness` so skills can trace back to their proof.

```bash
# Ingest LinkedIn profile (save page HTML or use Playwright MCP to capture)
uv run python .claude/skills/jobhunt/jobhunt.py ingest-job \
    --url "https://linkedin.com/in/username"

# Ingest resume (PDF or text)
# Upload or paste resume content, agent stores as jhunt-resume artifact

# Ingest ORCID profile
# Agent fetches via WebFetch and stores key publications/experience

# Ingest Google Scholar page
# Agent fetches and extracts publication list, h-index, citation metrics
```

The agent should fetch each source, extract the content, and store it as an artifact linked to the seeker role via `alh-aboutness`. This creates the evidence chain: artifact → seeker → skill.

#### Step 2: Sensemaking Interview

The agent reads the ingested artifacts and conducts a structured interview with the seeker:
- "Tell me about your TypeDB experience — what have you built?"
- "How deep is your AWS knowledge — have you architected solutions or followed tutorials?"
- "What's your biomedical ontology experience — which ones have you used?"

For each skill identified, the agent assesses the proficiency level based on the evidence and discussion.

#### Step 3: Create Skills with Evidence

```bash

From sciknow-io/alhazen-skill-examples, MIT.

More from sciknow-io/alhazen-skill-examples