indeed-brightdata（确实是光明数据）

Indeed Bright Data Skill

Search Indeed for job listings and company info via Bright Data's Web Scraper API. Designed for recruiting workflows on messaging platforms (Telegram, Signal) with smart defaults.

Prerequisites

• BRIGHTDATA_API_KEY environment variable must be set
• curl and jq must be available

Workflow Decision Tree

User wants job info?
├── Has a specific Indeed URL?
│   ├── Job URL (/viewjob?) → indeed_jobs_by_url.sh [SYNC — seconds]
│   ├── Company jobs URL (/cmp/*/jobs) → indeed_jobs_by_company.sh [ASYNC — minutes]
│   └── Company page URL (/cmp/*) → indeed_company_by_url.sh [SYNC — seconds]
├── Wants to search by keyword/location?
│   └── indeed_smart_search.sh [ASYNC — 3-8 min]
│       Agent says: "Searching now, this takes a few minutes."
│       If results < 5: auto-expands date range, do NOT ask user
│       Always pipe output through: indeed_format_results.sh --top 5
├── Wants company info?
│   ├── Has Indeed company URL → indeed_company_by_url.sh [SYNC — seconds]
│   ├── Has keyword → indeed_company_by_keyword.sh [ASYNC — minutes]
│   └── Has industry + state → indeed_company_by_industry.sh [ASYNC — minutes]
└── Check pending results? → indeed_check_pending.sh (run on heartbeat)

Always prefer sync (URL-based) scripts when the user provides a URL — they return in seconds.

Scripts Reference

Quick Start

User says: "Find me cybersecurity jobs in New York"

scripts/indeed_smart_search.sh "cybersecurity" US "New York, NY" \
  | scripts/indeed_format_results.sh --type jobs --top 5

User says: "Get details on this job: https://www.indeed.com/viewjob?jk=abc123"

scripts/indeed_jobs_by_url.sh "https://www.indeed.com/viewjob?jk=abc123"

Behavior Rules (MANDATORY)

1. NEVER return raw JSON to the user. Always pipe results through indeed_format_results.sh.
2. NEVER ask "want me to try broader keywords?" if results < 5. The smart search auto-expands automatically. Just tell the user: "Found only N results with recent postings, expanding search..."
3. NEVER present results older than 30 days without noting they may be stale.
4. When a discovery search is running, immediately acknowledge: "Searching Indeed now — this usually takes 3-5 minutes. I'll come back with results."
5. If the user asks a follow-up while a search is pending, run indeed_check_pending.sh first before starting a new search.
6. For Telegram: keep each message under 3500 characters. Use the ---SPLIT--- markers from indeed_format_results.sh to break across messages.
7. Always show total result count and offer to show more: "Showing top 5 of 23 results. Want to see more, or filter by salary/location?"
8. Default to "Last 7 days" for date filtering. If the user says "find me jobs" without a time preference, the default is already set.

Smart Search (Primary Entry Point)

# Basic search (expands keywords, deduplicates, defaults to last 7 days)
scripts/indeed_smart_search.sh "cybersecurity" US "Remote"

# All-time search
scripts/indeed_smart_search.sh "nursing" US "Texas" --all-time

# Skip keyword expansion
scripts/indeed_smart_search.sh "registered nurse" US "Ohio" --no-expand

# Bypass 6-hour cache
scripts/indeed_smart_search.sh "data science" US "New York" --force

Output is {"meta": {...}, "results": [...]} with metadata including query params, keywords used, and result counts.

Result Formatting

# Telegram-friendly summary (default)
scripts/indeed_format_results.sh --type jobs --top 5 results.json

# CSV export
scripts/indeed_format_results.sh --type jobs --format csv results.json

# Companies
scripts/indeed_format_results.sh --type companies --top 5 companies.json

# Pipe from smart search
scripts/indeed_smart_search.sh "nurse" US "Ohio" | scripts/indeed_format_results.sh --top 5

Heartbeat: Checking Pending Results

scripts/indeed_check_pending.sh
# Output: {"completed":[...],"still_pending":[...],"failed":[...]}

Run this periodically. If ~/.config/indeed-brightdata/pending.json exists and is non-empty, check for completed results. Format completed results with indeed_format_results.sh and send to the user.

Exit Codes

Caching

Smart search caches results for 6 hours. Identical searches (same keyword + location + country) return cached results without API calls. Use --force to bypass. Old results (>7 days) are auto-cleaned by indeed_check_pending.sh.

Data Storage

All persistent data is stored under ~/.config/indeed-brightdata/:

Auto-cleanup runs at the start of indeed_check_pending.sh. No data is sent anywhere other than the Bright Data API.

Security

All scripts source scripts/_lib.sh for shared HTTP and persistence functions. The library:

• Makes requests to a single endpoint: https://api.brightdata.com/datasets/v3
• Uses one credential: BRIGHTDATA_API_KEY (sent via Authorization: Bearer header)
• Writes only to ~/.config/indeed-brightdata/ (see Data Storage above)
• Does not read other environment variables, contact other hosts, or modify files outside its config directory

For full API parameter details

See references/api-reference.md for complete endpoint documentation, response schemas, and country/domain mappings.

For keyword expansions

See references/keyword-expansions.json for the lookup table of keyword-to-job-title mappings.