/tools/spam-score scores a draft email against ~10 SpamAssassin-style heuristics, in real time as you type. No backend, no upload — your subject and body are analysed by the React island in your tab.

Rules covered:

• Empty subject (+2.0) — aggressively flagged everywhere.
• Subject over 78 chars (+1.5) — most mailers clip there.
• Subject mostly UPPERCASE (+2.5) — > 50 % caps.
• Multiple ! in the subject (+1.5).
• Faked Re: / Fwd: without context (+1.0) — filters cross-check with your sent history.
• Body > 30 % UPPERCASE (+2.0).
• High link density — more than 5 links AND > 8 % of words (+2.0); just over 10 links anywhere (+1.0).
• ~80 spam-trigger phrases scored cumulatively up to +4.0 (free, winner, guarantee, act now, viagra, the classics).
• Very short body with links (+1.5) — phishing pattern.
• Three or more $ (+1.0).
• A small credit for bodies with no links at all (−0.5).

Total is bucketed into < 2 = Likely Inbox (teal), 2-5 = Borderline (amber), 5+ = Likely Spam (red). Each rule that fires is rendered as its own row with the +score, the label and the explanation — so you know exactly which line to tweak.

This is a fast iteration loop for a transactional / newsletter draft. It is not a real Gmail / Outlook delivery test — for that you need actual seed inboxes, which is what the upcoming inbox-placement tool will do once we wire up our first Gmail seed (see FerrLens-Cloud#44).