Token Safety Screen

Checks a token's creator for a wallet-reusing serial-rug track record, before an agent buys.

An agent about to buy a freshly-launched token can't tell a real project from the 71st throwaway coin of a wallet that rugs on a schedule. The Safety Screen answers that in one call, from on-chain evidence: it resolves the token's creator from the immutable on-chain record, walks the creator's prior launches, and asks whether that history is a pattern of launching and abandoning. Payfetch can consult it before an agent pays for a freshly-launched asset; it also stands alone as a single x402 call.

Read the full methodology →

The API

POST /v1/safety/screen        # basic screen: $0.01
POST /v1/safety/screen/deep   # deep screen:  $0.05
{ "mint": "<token address>", "chain": "solana" }

Basic returns the topline verdict + score. Deep adds the re-judgeable deployer-cluster evidence block (labeled/abandoned counts, abandonment rate, cluster id, coverage, and the frozen thresholds version).

Verdict	Meaning
`danger`	The creator cluster has a confirmed serial launch-and-abandon record.
`caution`	Mixed or partial signals worth a human look.
`safe`	A sustained token / no adverse deployer record found.
`unknown`	Not enough history to judge. We say so rather than guess.

Real footage: a Claude Code agent pays $0.05 for the deep screen before buying a pump.fun token — the deployer's launch-and-abandon record comes back with raw counts, settled on Base (tx hash in frame).

Measured accuracy

Our frozen, labeled evaluation set holds 31 tokens from wallet-reusing serial deployers and 54 control tokens from legitimate or established creators. On it, screen version p1s-detect-1.3.1 flags 31 of 31 serial-deployer tokens, with 0 false-danger verdicts on the control set and 0 false-safe verdicts overall. The evaluation is point-in-time (no lookahead) and re-runnable; the labeled set and thresholds version are frozen so the number can't quietly move.

The claim is scoped, and we state the scope: the screen detects wallet-reusing serial creators, deployers whose launch-and-abandon history is visible because they reuse wallets. A creator who rotates to a fresh wallet for every launch is out of scope for v1; extending coverage to rotation clusters is tracked, deferred work, and we will not quote an "all serial ruggers" number until it exists.

How verdicts are made, and their limits

Track record, not intent. Verdicts measure a deployer cluster's history of launching tokens that never graduated and went dead (an observable, on-chain fact), not a prediction of intent.
Honest uncertainty. A creator whose prior launches are all too young to have an outcome yet returns unknown (insufficient_history), never a premature danger. The screen would rather admit it can't tell than be confidently wrong.
No lookahead. A judgment is point-in-time: it uses only what was on-chain as of the query, so a label can't be contaminated by the future.
Independent ground truth. Outcomes are derived from raw on-chain signatures, balances, and the launchpad's own indexer, never from another risk vendor's flag graded against itself.
Degrade toward uncertainty. If an upstream is unavailable or a program layout is unrecognized, the screen returns unknown. It never emits a confident verdict it can't stand behind.
Coverage (v1): Solana pump.fun launches. More launchpads and chains follow the same append-only, reproducible methodology.

Forum Labs · Payfetch · Trust Score · Token Safety · Methodology · GitHub · ops@forum-labs.com · @shopforumlabs
© 2026 Forum Labs. Informational screening from public on-chain data. Not financial or investment advice; not a guarantee about any token.