B2B Lead Agent

Run the bundled CLI from this directory with the current platform's command executor. If commands cannot run, stop and report the missing permission. Use python or python3; do not name platform-private tools, Gmail, or a fixed browser.

Guardrails

• Public websites/contact data only.
• No captcha/login/paywall/rate-limit bypass, proxy pools, forged fingerprints, LinkedIn scraping/auto-connect, or SMTP probing.
• Keep evidence text + source URL for conclusions, scoring, and personalization.
• Guessed emails are never valid or sendable.
• Real send requires one explicit lead_id plus a second confirmation; default mode is draft-only.

CLI

Every command returns JSON. Treat success:false or non-empty errors as review-needed.

python -m scripts.cli init
python -m scripts.cli search --config config.yaml --input-csv data/companies.csv
python -m scripts.cli <scan|score|draft> --config config.yaml
python -m scripts.cli <export|status|retry-errors>
python -m scripts.cli approve-send --lead-id LEAD_ID [--confirm]

Flow

1. init, fill config.yaml, then provide CSV columns company_name,website,country,source_url.
2. Run search -> scan -> score -> draft -> export; use status between steps.
3. Use approve-send only for a user-specified lead_id; first call previews, --confirm records approval in draft-only mode.

Reference Routing

• Workflow/recovery: references/workflow.md
• Scoring changes: references/scoring-rules.md
• New sources, browser automation, send adapters: references/compliance.md
• Draft/approval changes: references/email-rules.md
• Failures/locked exports: references/troubleshooting.md