Automation policy

Polite by design — and we mean it measurably.

"Automation = spam" is a fair worry, so we made restraint a product rule instead of a promise. Every directory in the catalog is classified by how far automation may politely go, and the agent's behaviour is bounded by the rules below. Nothing here is aspirational — it's how the runner actually works.

The rules

  • Never bypass CAPTCHAs or anti-bot protections.
  • Never fake engagement — no fake votes, comments, reviews, or upvotes.
  • Never submit a product to an irrelevant directory to pad a number.
  • Never post community content without the customer's explicit approval.
  • Never claim a customer's identity without authorization.
  • Never resubmit after a rejection without correcting what was wrong.
  • Respect rate limits, robots.txt, and each platform's terms.
  • The customer's accounts and credentials stay the customer's — we act with consent, never ownership.
  • Keep screenshot evidence of every submission — the form and the result.
  • Stop an automation that fails twice and hand it to a guided human flow.
  • Always provide a manual fallback (the Tier-3 guided tutorial).

How every directory is classified

Each of our 1,058 active directories carries a stored automation classification, derived automatically from its real signals (terms, captcha, account requirements, editorial review, our automation tier). Live counts:

Browser automation201

A public form with no captcha and no account. The agent fills and submits it end-to-end, then screenshots the result.

Assisted, with consent238

Needs your account, login, or email verification. The agent acts on your behalf only with your explicit consent, using your own credentials.

Editorial review289

A human editor decides. We submit a tailored, relevant listing; whether it's accepted is their call, and we never fake engagement to influence it.

Human-guided329

A captcha, anti-bot wall, or community mechanic. We do NOT bypass it — the agent pre-fills a guided, step-by-step tutorial and hands it to you.

Automation prohibited1

The site's terms forbid automated submission. We never automate these — they're guided-only or skipped.

Auditable metrics

A boundary you can't measure is a slogan. These are the numbers we hold ourselves to — the ones marked enforced are guaranteed by how the code is built; the rest are being wired into a live measurement dashboard.

CAPTCHA-bypass attempts0enforced

The agent has no captcha-solving capability; captcha detection routes to a guided human flow instead.

Auto-stop on repeated failure2-strikeenforced

Two consecutive automation failures demote the submission to a guided manual flow — no infinite retries.

Paid-fee submissions auto-claimed as done0enforced

A paywall is detected and the submission is marked payment-required, never falsely reported as live.

Consent captured before assisted actions100%instrumented next

Assisted (account/login) directories require the customer's credentials and consent; measured on the dashboard (VIR-570).

Irrelevant-submission rate→ 0instrumented next

Relevance matching (fit scoring) gates which directories a product is submitted to; tracked on the dashboard (VIR-570).