tick

name	Tick
description	Run one AUTONOMOPOLY agent tick — claim fees, LP DIEM, LP range check + reposition, maintenance inference
var
tags	["agent","on-chain","defi"]

Run the agent tick. Execute:

node --import tsx scripts/queue-intent.ts tick

Check memory/goals.json for the current mode before proceeding. The harness injects it as AGENT_MODE but read the file directly to confirm.

Every tick (accumulate + build mode)

1. Reads claimable DIEM from FeeLocker 0xF7d3BE3FC0de76fA5550C29A8F6fa53667B876FF
2. Claims if ≥ 0.1 DIEM
3. Reads wallet DIEM balance
4. LPs into ETH/DIEM Uniswap v3 1% pool if ≥ 0.1 DIEM
5. Otherwise runs maintenance inference via Venice llama (free tier)

Agent wallet: 0x8767Df39eCeeaeB11554642237aC4E08660aB6A3

If the tick fails with a Privy error, log the full error to memory/logs/${today}.md and send a notification via ./notify with the error summary.

If the tick succeeds, log a one-liner to memory/logs/${today}.md:

tick: claimed Xm DIEM, LP'd Y DIEM | ticks=[A,B] currentTick=C

or

tick: nothing to claim/LP | maintenance inference ran

Every tick — LP range check (absorbed from lp-monitor, 2026-06-10)

The standalone lp-monitor skill is disabled — its duties run here so one job per 2h covers both (same chain reads, half the runs).

1. Read all positions:

   node --import tsx scripts/check-portfolio.ts
   

   Record every tokenId, in-range vs out-of-range, current pool tick, FeeLocker balance, wallet balances.

2. In range → note fee accrual in the log line. Out of range → queue a reposition intent for the gated executor (you cannot sign in this step):

   node --import tsx scripts/queue-intent.ts reposition --token-id <tokenId>
   

   The executor runs scripts/reposition.ts (reads liquidity on-chain, claims FeeLocker fees, swaps 50%, mints a new in-range position, records the new tokenId in memory/lp-positions.jsonl).

3. Safety checks before queueing: dry-run shows the right tokenId with non-zero liquidity; the new range brackets the current tick; ETH > 0.003 for gas. If a check fails, log the issue and notify instead of queueing.

4. Log to memory/logs/${today}.md:

   lp-check: N in range, M reposition queued | tick=C | FeeLocker=X DIEM
   

   Notify via ./notify only when a reposition was queued or a safety check failed — all-in-range is log-only.

After every tick — queue tweet content

Read memory/goals.json. If tweetingPaused is true, skip this entire section — do not write any files to .pending-x/. Log "tick: tweet generation skipped (tweetingPaused=true)" to memory/logs/{today}.md.

After the tick runs, generate tweet files in .pending-x/.

In accumulate mode: generate 2 tweets (see below). In build mode: generate 3 tweets — add a build-update tweet (see below).

Filename format: tweet-{YYYYMMDD-HHMMSSsss}-{type}.txt — use seconds and milliseconds (e.g. tweet-20260609-143022847-on-chain-report.txt). This prevents same-basename collisions when two tweets are generated in the same second. Generate the timestamp once at the start of this section and reuse it for all files in this run.

Non-empty body assertion: Before writing any file, assert that the tweet body (everything after the #content_type: line) is non-empty and under 280 characters. If the body is empty, log an error to memory/logs/{today}.md and skip that file — do not write an empty stub.

1. Self-report (on-chain-report or lp-update): one tweet about AUTONO's current state — balance, mode, LP position, or fee rate. Use real numbers from the tick output. File: tweet-{YYYYMMDD-HHMMSSsss}-on-chain-report.txt

2. Outward signal — read the ## Content Type Weights table from memory/x-strategy.md. Select a content type using the weights as probabilities, excluding on-chain-report and lp-update (those are self-report types). If the file is absent or all relevant weights are insufficient_data, use equal weights across: base_signal, ecosystem_reaction, agent-philosophy, contrarian, question, milestone. Then web-search at least one source from ## External signal sources in memory/x-strategy.md. Pick the most interesting result from the last 48h and write a tweet with a specific take — not generic commentary. If no compelling signal is found, fall back to agent-philosophy (one concrete belief from identity/SOUL.md, stated plainly in one sentence). File: tweet-{YYYYMMDD-HHMMSSsss}-{type}.txt

Each file format:

#content_type:{type}
{tweet text — no hashtags, no emojis, under 280 chars}

3. (Build mode only) Build-update (build-update): one tweet about what you are actively building or what you just improved. Must include a specific detail — a file name, a metric, a before/after. No vague "I'm building the launchpad." Example: "self-improve ran: increased contrarian weight 0.15→0.22 based on 3-week engagement data. top-3 performing tweets all contrarian. feedback loop closing." File: tweet-{YYYYMMDD-HHMMSSsss}-build-update.txt

Do not repeat a topic covered in memory/x-tweet-log.jsonl within the last 48h. Check before writing.

After every tick — Dependabot check

After the tick completes (success or failure), run:

gh pr list --author app/dependabot --state open --json number,title,createdAt,url 2>/dev/null

Include any open Dependabot PRs in the ./notify message. Format:

tick: <summary> | Dependabot: N open PR(s): #X title1, #Y title2

or omit the Dependabot section entirely if N=0.