pump-fun-compose

Agent-level loops that compose pump-fun (read) and pump-fun-trade (sign).

Tool	Loop
researchAndBuy	search → vet (curve, holders, creator) → quote → buy
autoSnipe	poll new tokens → vet → buy until session cap
copyTrade	watch a wallet's recent buys → mirror with size scaling
rugExitWatch	watch held mints → sell on concentration or dev-sell triggers

Filters (manifest.config)

• minHoldersForBuy — reject thin books
• maxTopHolderPct — reject whale-controlled supplies
• rejectIfCreatorRugCount — reject repeat ruggers
• exitOnConcentrationPct / exitOnDevSellPct — exit triggers
• sessionSpendCapSol / perTradeSol — hard spend caps

Args supported by every loop

• dryRun: true — vet/quote runs, buyToken / sellToken are skipped and sig: null is returned. Spend cap still ticks so a dry-run terminates the same way a live one would. Use it to validate filters before going live.
• sessionId: "<stable-string>" — persists seen / mirrored / spent / exited to ctx.memory (in-process Map fallback). Reusing the same id resumes a prior session — preventing a crash mid-loop from re-spending up to the cap.
• signal: AbortSignal — every poll, sleep, inner-loop, and outer-loop checks signal.aborted. Stop is responsive (≤ ~1 RPC call) regardless of pollMs.
• onProgress(evt) — called on every state change with { type, mint?, sig?, spent?, reason? }. Drive a live UI off it.

The result data.reason indicates why the loop exited: 'aborted' | 'duration' | 'cap' | 'all-exited'.

Wiring

The host runtime injects:

• ctx.skills.invoke('skill.tool', args) — sibling-skill calls
• ctx.wallet — used by the trade skill (this one never touches keys directly)
• ctx.memory.note — every action is auditable

Composition example

await ctx.skills.invoke('pump-fun-compose.autoSnipe', { durationSec: 600 });
// then babysit positions
await ctx.skills.invoke('pump-fun-compose.rugExitWatch', {
  mints: heldMints,
  durationSec: 3600,
});