{"id":"1d83688c-de01-40b4-abe7-f58c24a8188f","slug":"solana-token-safety-skill","kind":"skill","name":"solana-token-safety-skill","summary":"Pre-trade / pre-launch token safety for Solana agents — the market-data-integrity layer contract checkers miss (bot/wash volume, fake market cap, price bait, single-wallet float). A skill for the Sola","owner":"ernanibmurtinho","repoUrl":"https://github.com/ernanibmurtinho/solana-token-safety-skill","license":"MIT","licenseTier":"permissive","trust":"community","stars":0,"match":null,"fewStars":true,"format":"skill","createdAt":"2026-06-19T03:40:05.000Z","addedAt":"2026-10-21T13:20:03.631Z","logoUrl":"https://avatars.githubusercontent.com/u/9011171?u=094c1a43ebe4e48d55cce65f188f277f10b7da9f&v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["data-analytics"],"tags":["ai-agents","anti-rug","mev","solana","solana-ai-kit","token-safety"],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-23T19:59:13.712Z","sample":false,"url":"https://conduits.dev/tools/solana-token-safety-skill","language":"Python","transports":["stdio"],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# solana-token-safety-skill\n\n**A second opinion on whether a Solana token's *market* is real — before an agent\ntrades it or a builder lists it.**\n\nBuilt for the [Solana AI Kit](https://github.com/solanabr). A progressive,\ntoken-efficient skill that gives any coding/trading agent a one-call pre-trade\nsafety gate.\n\n## Try it in 30 seconds\n\nNo key, no install — hit the live oracle:\n\n```bash\ncurl -s -X POST https://api.geckovision.tech/safety \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\"mint\":\"JUPyiwrYJFskUPiHa7hkeR8VUtAeFoSYbKedZNsDvCN\"}'\n# -> {\"gate\":\"ok\", ... \"information_mev\":{\"label\":\"clean\", ...}}\n```\n\n…or run the zero-dependency reference client (prints the verdict, exits non-zero\non a block so it composes in CI):\n\n```bash\npython scripts/safety.py JUPyiwrYJFskUPiHa7hkeR8VUtAeFoSYbKedZNsDvCN\n# ✅  gate = OK  (JUPyiwr…)\n#     - No manipulation signals: liquidity 0.40% of market cap.\n```\n\n## The problem it solves\n\nEvery contract checker on Solana (RugCheck, GoPlus, Solsniffer, Bubblemaps) tells\nyou whether the **contract** is safe — mint/freeze authority, honeypot bytecode.\nThey routinely **pass tokens whose price and volume are manufactured**: bot-\ninflated wash volume, a fake market cap held up by a few thousand dollars of\nliquidity, the same token quoted far above its real price in thin satellite\npools, one wallet sitting on the float.\n\nThat's the gap that drains agents and rugs retail. A static contract scan can't\nsee it, because the manipulation is in the **market data**, not the contract.\n\nThis skill is that missing layer. It flagged a live token (BrCA) as\n`block / manipulated 0.95` — $5.73M quoted market cap on **$6.5K** of liquidity,\ntop holder 72% — while the platform it traded on rated it \"Normal\".\n\n## What it does\n\nOne call → a one-glance gate:\n\n```\nblock    — do not trade / do not list (fake mcap, manipulated/wash, depeg, honeypot)\ncaution  — reduced size + explicit confirm (thin liquidity, concentration, elevated)\nok       — checked and clean\nunknown  — the check could not run (fail-OPEN — never assume safe)\n```\n\n…plus the `reasons[]` to explain the verdict to a human, and (as the Launch-\nFirewall stream rolls out) a `wash_risk` block with the specific flow signal that\nfired (one-sided buy-loop, multi-pool price bait, wash recirculation, sybil\nfunder cluster).\n\nIt's **free, keyless, and fail-OPEN** — `POST` a mint, get a gate.\n\n## Install\n\n```bash\ngit clone <this-repo> && cd solana-token-safety-skill\nbash install.sh          # copies the skill + /safety-check command into ~/.claude\n```\n\nOr point it at a custom config dir: `CLAUDE_DIR=/path bash install.sh`.\n\n## Use\n\n- Ask your agent: *\"is this token safe to trade: `<mint>`?\"*\n- Trader command: `/safety-check <mint>` · Issuer command: `/launch-check <mint>`\n- Agent persona: `token-safety-auditor` runs the full diligence flow\n- Wire it into an agent's pre-trade loop → [`skill/agent-integration.md`](skill/agent-integration.md)\n\n## Two sides, one check\n\n- **Agents** — gate a buy/swap before you sign (`/safety-check`, the `pre-trade-safety` rule).\n- **Issuers / launchpads** — vet a token's launch for bot manipulation in the first\n  candles (`/launch-check`, the Block-Zero read).\n\n## Structure\n\n```\nskill/\n  SKILL.md                 entry point + routing (load this)\n  quickstart.md            the one call (curl / Python / TS)\n  signals.md               what each flag/signal means\n  interpreting-verdicts.md gate semantics + fail-OPEN + FP awareness\n  agent-integration.md     MCP + HTTP pre-trade gate pattern\n  launch-check.md          issuer / Block-Zero read\ncommands/\n  safety-check.md          /safety-check <mint>   (trader)\n  launch-check.md          /launch-check <mint>   (issuer)\nagents/\n  token-safety-auditor.md  market-integrity specialist persona\nrules/\n  pre-trade-safety.md      opt-in: gate token swaps with a check first\nexamples/\n  clean-jup.json           real OK verdict (verified) + blocked-brca.json (illustrative)\nscripts/\n  safety.py                zero-dependency reference client (stdlib urllib)\n  safety.ts                zero-dependency reference client (built-in fetch)\ninstall.sh                 copy-only installer (no binaries, no network)\nCLAUDE.md · LICENSE (MIT) · README.md\n```\n\n## How it works\n\nThe skill is a thin, safe client over **Gecko's free `/safety` oracle**\n(`POST https://api.geckovision.tech/safety`) or the `gecko_safety` MCP tool. The\noracle reads on-chain liquidity/holder/market data, computes deterministic\nmanipulation signals, and returns the gate. The skill ships **no executable\nlogic of its own** beyond the install copy — nothing opaque, nothing to audit\nbeyond markdown.\n\n- Docs: https://docs.geckovision.tech\n- Provider: [Gecko](https://geckovision.tech) — a decision-integrity oracle for\n  Solana agents. The `/safety` tier is free by design; the considered multi-voice\n  adjudication lives behind the paid `/trade_research` tier (not used by this skill).\n\n## License\n\nMIT — ready to be merged or submoduled into the kit.\n"}