---
name: trench-convergence-trading
description: Screen, score, size and manage memecoin trades on Robinhood Chain by reading the Trench Terminal web app (trenchterminal.app) in a browser — its convergence board, coin pages, wallet leaderboard and clusters. Use this skill whenever an agent is asked to find trades, evaluate a coin, decide entries or exits, manage open positions, or interpret KOL convergence, hold ratio, hold age, clusters, ticket sizes, launchpad vs graduated coins, or PRE-TAPE sellers on Trench Terminal — even if the request just says "what should I buy", "check this coin" or "manage my bags".
---

# Trench Convergence Trading

A rules-based playbook for trading off what tracked KOL and smart-money wallets are
doing, read from the Trench Terminal screens. It is a framework, not financial
advice, and it has not been backtested: treat every threshold as a starting point
to be tuned from the trade journal.

Most memecoin trades lose, so the winners have to pay for them. The rules exist to
keep losses small and uniform, and to skip setups that look like conviction but
are really exit liquidity.

## Authority

- **Default: recommend only.** Output decisions in the format at the end of this
  skill. Execute a swap only if the operator has explicitly authorized execution in
  this session, with a wallet and limits given.
- **Hard limits are not negotiable** by the agent, by a coin's setup, or by a
  winning or losing streak: position size, max open positions, daily stop,
  stop-loss. Only the operator can change them.
- **Read the page, not the network.** Use only what the terminal renders on screen.
  Do not call the site's internal endpoints or scrape its network traffic.
- **Missing data means skip, not guess.** A value shown as `-`, or one you cannot
  read, fails the rule that needs it.

## Reading the terminal

### Navigation

- **Switch to Pro view first.** The terminal opens in Simple view (cards with the
  key numbers). Click `Pro` at the top right or press `P`; the choice is
  remembered. Everything below describes Pro view, which shows every column.
- **Search / command line** (top, `TRENCH>`): type and press Enter.
  - `NEO` or a contract address opens that coin. If several contracts share the
    ticker, a picker lists them, most liquid first.
  - `CONV 1H`, `CONV 15M`, `CONV 6H` … opens the convergence board for a window.
  - `TLDR` opens the coins that pass every rule in this skill's entry checklist.
  - `WLT <handle>` or a wallet address opens one wallet.
  - `WALLETS`, `FLOW 1H`, `TAPE`, `TAPE NEO`, `HELP`, `PRO`, `SIMPLE`.
- **Keys:** `0` TLDR, `1` board, `2` coin, `3` wallets, `4` flow, `5` tape; `P`
  Simple/Pro; `[` and `]` change the window; arrows move the cursor; Enter opens
  the row under it; Esc goes up a level.
- **Window chips** (top right: `15M 1H 6H 24H 72H`) change the board, flow and the
  coin chart's range together.
- **Direct URLs:** `#tldr`, `#conv/1H`, `#coin/<address>`, `#wallet/<address>`, `#wallets`.
- **Status pill** (top): `Live` with the chain's head block. `streaming` means
  trades are pushed as blocks land; `in sync, polling 2s` means the terminal reads
  the chain every 2 s. `Stalled` means no update for a while: treat the board as stale.

### Number formats

- `$1.2K` = 1,200; `$3.4M` = 3,400,000; `$1.1B`, `$2.0T` likewise.
- `-` means no data (treat as null).
- `~` before an ATH (`~$1.2M`) means the ATH is only the highest price seen since
  tracking began, not the coin's full history. Treat distance-from-ATH rules as
  weaker when you see it.
- Ages such as `12m`, `3h05`, `2d4h` are time since the event.

### TLDR (`TLDR`, key `0`)

The terminal applies this skill's entry checklist to the 1H board once for every
user and lists the coins that pass: at least 6 of the 12 checks and none of the
automatic skips. Pro columns: TOKEN, PATTERN, SCORE (checks passed, with one pip per
check in order 1–12), HOLD (effective holders / entered), RATIO, OLDEST, CHART
(alive, dead, unclear), X, MC, LIQ, SIZE, STOP, TIME. Selecting a row shows the
entry floor and which checks it missed. Launchpad coins are marked `CURVE` and
rank 1.5 checks lower than an otherwise equal listed coin. Use TLDR as the
candidate list, then verify each candidate on its coin page with the rules below.

### Convergence board (`CONV <window>`)

Header strip: money in, money out, net, wallets, coins, trades for the window, and
a flow bar chart (green up = USD bought, red down = USD sold, white line =
cumulative net).

Table columns, per coin:

| Column | Meaning |
|---|---|
| TOKEN | Ticker, then `CA` (copy contract address) and `DS` (open DexScreener). `CLONE` tag = another contract with this ticker is more liquid. `x3` = three contracts share it and this one is the most liquid. `LP` = launchpad: no market cap yet. |
| STATE | `Heating` new tracked buys in 15 min and most buyers holding; `Holding`; `Stalled` little real trading (or no tracked buys for 30 min on a launchpad coin); `Exiting` most tracked buyers have sold |
| FLG | `C` clone, `L` thin liquidity, `N` pair under 1 h old, `D` more than 80% below ATH |
| KOLS | **Entered**: distinct tracked wallets that bought in the window |
| IN | One block per wallet (max 12, then `+N`): bright green = first-ever buy and holding, green = holding, orange = trimmed more than 5%, hollow = sold out |
| HOLD | **Holding**: of those, still holding |
| EXIT | **Exited**: sold in the window and no longer holding |
| NET | USD bought minus sold by tracked wallets in the window |
| 1ST IN | First tracked buyer in the window and how long ago |
| ENTRY | USD-weighted average market cap at which they bought |
| MC | Market cap now |
| X | MC ÷ ENTRY |
| ATH MC, ATH% | All-time-high market cap and distance from it |
| LIQ, 1H | Liquidity, 1 h price change |
| SCORE | The ranking score. Rows are sorted by it. A coin every buyer has left scores 0 |
| 24H | 24 h sparkline |

Ranking: each buyer counts by how much of its position it still holds (a wallet
that sold out keeps a tenth of its weight), recent and first-ever buys count more,
each wallet that sold out subtracts, and the total is multiplied by the hold ratio
and by live trading on the coin (volume against market cap, transaction count, buy
pressure). Launchpad coins, which have no market data, carry a mild discount.

Selecting a row expands a detail line listing each buyer: handle, time since their
first buy **in this window**, USD bought, `1ST` (first-ever buy), `OUT` (exited),
`-40%` (trimmed that much).

### Coin page (`COIN` / open a row)

- **Header card:** ticker, name, flags, full contract address, market cap and 1 h
  change, and buttons: Copy CA, DexScreener, GeckoTerminal, Trades on the tape.
  `Launchpad` and `MC unknown` mean DexScreener has no market data yet. A red box
  under it means a more liquid contract uses the same ticker.
- **Key numbers:** KOLs holding (`3 of 5`), Holders' entry → now (the X of current
  holders), KOL net flow, Liquidity, Volume 1h, From ATH.
- **Chart:** the market-cap line; green up-triangles are tracked buys, red
  down-triangles are tracked sells, a dashed red line is the ATH. Hovering shows the
  time and market cap, or the trade under the cursor. Switch the window chip to see
  a longer or shorter range.
- **All stats** (open in Pro view): MC, ATH MC, From ATH, ATH set, Price, Liquidity,
  Liq / MC, Pair age, Vol 1h, Vol 24h, Txns 1h buy/sell, Txns 24h buy/sell, 5m, 1h,
  6h, 24h, KOLs in, KOLs holding, Holder entry, Holder x, KOL net flow, Quoted.
- **Tracked wallets table:** WALLET (`1ST` tag = first-ever buy), STATUS (`Holding`,
  `Trimmed`, `Sold out`, `Pre-tape`), PNL, X, FIRST IN (**hold age**: time since
  that wallet's first buy of this coin), ENTRY MC, COST, SOLD. This table is
  complete; use it instead of the board's capped buyer list.
- **Trades table:** every tracked trade with time, wallet, side, USD, MC AT (market
  cap at that trade's price) and, in Pro view, amount.

### Wallets screen (`WALLETS`)

- **Leaderboard:** WALLET, PNL, WIN, OPEN, TRD, TOK, IN $, REAL, UNREAL, SCOUT.
  SCOUT `3/4` = among the first three tracked wallets into 3 of the 4 tokens that
  3+ tracked wallets later bought. Click a header to sort.
- **Clusters:** a heatmap of shared tokens, and a pairs table: LEADS, FOLLOWS,
  SHR (tokens both bought), SYNC (shared tokens both entered within 60 s), LAG
  (median seconds apart), J (overlap). SYNC shown in orange and equal to SHR means
  every shared entry was simultaneous: usually one operator or a copy bot.

### Wallet page (`WLT <handle>`)

Key numbers (PnL, open positions, open value, realized, unrealized, first buys), a
POSITIONS table (token, status, PnL, X, MC now, first in, entry MC, cost, sold) and
a TRADES table with USD per trade. Use Buy rows to judge the wallet's usual ticket.

### Known limits

- **Latency:** with streaming on, trades reach the screen within a few seconds of
  their block; when polling, within about 2–5 s. You still enter after the wallets
  you follow. The page updates itself; never reload to refresh.
- **Tape only:** PnL and holdings reflect trades since tracking began. A wallet that
  sold more than the tape saw it buy shows as `Pre-tape`.
- **Airdrops are filtered:** tokens received without payment never appear as buys.
  Selling them does appear, as `Pre-tape`. See "Dumped allocations".
- **Launchpad coins** show `LP` / `Launchpad`, `-` for MC, LIQ and VOL, and have no
  market chart. Curve progress must be read on the launchpad itself. If you cannot
  read it there, skip curve coins.

## Definitions

Compute these for every candidate. Counts depend on the window selected.

- **Entered** = KOLS; **Holding** = HOLD; **Exited** = EXIT on the board row (Pro view).
- **Hold ratio** = HOLD ÷ KOLS.
- **Hold age** of a wallet = its FIRST IN on the coin page. Do not use the ages in
  the board's detail line; those count from the first buy inside the window only.
- **Oldest live holder** = the largest FIRST IN among `HOLDING` or `TRIMMED` rows.
- **Holder entry floor** = the smallest ENTRY MC among `HOLDING` or `TRIMMED` rows.
- **Dollars still held** = sum of COST for `HOLDING`/`TRIMMED` rows ÷ sum of COST
  for all rows (exclude `PRE-TAPE`).
- **Usual ticket** of a wallet = the median USD of BUY rows on its wallet page.
- **Acceleration** = the coin's KOLS in `15M` ≥ half its KOLS in `1H`, and VOL 1H
  higher than at your previous look.
- **Effective holders** = holders after merging clusters: two holders count as one
  when the CLUSTERS pairs table links them with SYNC ≥ 2 and SYNC equal to SHR.
- **Scout** = SCOUT on the Wallets screen of at least `3/4` (≥ 0.75 with 3+ tokens).

## Chart state: alive or dead

Hold age is meaningless without this. A wallet holding for hours on a dead chart is
a bagholder, not a believer. Read everything from the coin page.

**Alive** — all of:
- VOL 1H ÷ MC ≥ 0.05
- TXNS 1H B/S total ≥ 50, with buys at least 40% of it
- On the chart over the last 2 h, lows are holding or rising, **or** MC stays above
  the holder entry floor
- LIQ not down more than 20% since your previous look

**Dead** — any of:
- VOL 1H ÷ MC < 0.01
- TXNS 1H total < 20
- MC below the holder entry floor while the chart makes lower lows

Anything in between is **unclear**: treat it as not alive.

**Buy placement** on the chart:
- Green triangles in the lower third of the range, before the move: skilled entries.
- Green triangles stacked at the top of a spike: chasing, future exit liquidity.
- Red triangles at the highs while holders remain: healthy profit-taking. Red
  triangles from the 1ST IN wallet: warning.

## Patterns

Classify every candidate into exactly one pattern by comparing a short window
(`15M`/`1H`) with a long one (`6H`/`24H`) and the chart. The pattern sets size and
the time stop.

| Pattern | Short window | Long window / hold age | Chart | Action |
|---|---|---|---|---|
| **Second wave** | 3+ new entries | 3+ holders with hold age ≥ 2 h | alive, breaking a range high | Best setup. Full size. |
| **Seasoned conviction** | 1–2 new entries | 3+ holders with hold age ≥ 2 h | alive, higher lows | Full size. |
| **Fresh spike** | 3+ holders, all hold age < 15 min | none earlier | spiking, VOL 1H rising | Half size, 20 min time stop. |
| **Revival** | 3+ new entries | holders from hours ago | dead, VOL 1H rising sharply | Half size. The new entrants are the signal, not the old holders. |
| **Handoff** | 3+ new entries | earlier holders `EXITED` | topping, red triangles at the highs | Skip. |
| **Bags** | 0 new entries | holders from hours ago | dead | Skip. Not a signal. |

## Hold ratio

| KOLS | Hold ratio | Reading |
|---|---|---|
| 3+ | ≥ 0.8 | Accumulating, if the chart is alive |
| 3+ | 0.5–0.8 | Mixed. Acceptable only if the exited wallets are the most recent entrants |
| 3+ | < 0.5 | Distribution. Skip, or exit |

A high hold ratio on a dead chart means nobody bothered to sell. Weigh exits by hold
age: one seasoned holder exiting outweighs several 10-minute entrants.

## Ticket-size signals

Supporting evidence. They add checklist points only where the checklist says so.

- **Relative to the wallet's habits:** a buy ≥ 3× the wallet's usual ticket shows
  conviction. A buy far below it is a test or a bot.
- **Absolute:** $500+ from a named wallet outweighs many $20–50 buys.
- **Dollars still held** under 0.5 means the big money has left, even if the wallet
  hold ratio looks fine.
- **Big ticket + long hold + alive chart** is the strongest single-wallet signal. A
  big ticket on a dead chart is a bag.
- **Adding:** a holder with several BUY rows, the later ones at range lows, is strong.
  Adding into a spike is weak.
- **Relative to liquidity:** one tracked buy over 5% of LIQ has already moved the
  price. Wait for a pullback.
- **Combined:** $2K+ of COST still held in a pool under $50K LIQ is strong. The same
  in a $1M pool barely matters.
- **Sells:** SOLD of 20–30% (`TRIMMED`) into strength is normal. A large-COST wallet
  going `EXITED` counts double against the setup.

## Entry checklist

One point each. **Enter only at 7 or more.**

1. 3+ effective holders in the `1H` window
2. Hold ratio ≥ 0.8
3. Chart is alive
4. Oldest live holder ≥ 2 h on an alive chart, **or** the pattern is Fresh spike with rising VOL 1H
5. New entries in the last 15 minutes (the coin appears in `15M` with KOLS ≥ 1)
6. Tracked buys sit at range lows or before the move, not stacked at a spike top
7. The 1ST IN wallet shows `HOLDING` or `TRIMMED` on the coin page
8. At least half of the holders carry the `1ST` tag
9. NET positive, with at least $1K bought
10. X under 1.5
11. A scout among the holders, or a holder with positive PNL on the Wallets screen
12. At least one holder bought ≥ 3× their usual ticket, or dollars still held ≥ 0.5

### Automatic skips

Skip regardless of score if any is true:
- `C` in FLG, or a red clone line on the coin page
- `L` in FLG on a listed coin
- Hold ratio < 0.5
- Pattern is Handoff or Bags
- Chart is dead and the pattern is not Revival
- **Dumped allocations:** 2+ `PRE-TAPE` rows on the coin page
- Listed coin with LIQ under $20K or LIQ/MC under 5%
- FROM ATH worse than −60% unless the pattern is Revival
- A value a rule needs shows `-`

`N` in FLG (pair under an hour old): trade half size.

## Launchpad coins (bonding curve)

Recognized by `NOT LISTED ON DEXSCREENER` / `AWAITING QUOTE`, `-` market fields,
and confirmation on the launchpad. Judge on the coin page's TRACKED WALLETS table
(holders, hold ratio, hold age, COST) and the launchpad's curve progress.

- **Alive on a curve:** new curve buys every few minutes and progress moving.
  **Dead:** progress flat for 30+ minutes.
- Hold age on a curve counts only if the curve is still moving. Most curve buyers
  flip within minutes, so a live 1 h+ holder is rare and meaningful.
- **Enter** with 3+ effective holders, hold ratio ≥ 0.8, a moving curve, and at
  least one holder of 30+ minutes, when progress is 40–80%.
- **At graduation:** sell half into the graduation spike. Manage the rest as a
  graduated coin once the terminal shows market data for it.
- **Exit** if progress stalls 30 minutes or the hold ratio drops under 0.6.

## Sizing and risk

Assume a $100 starting bankroll; scale proportionally to the operator's balance.

- **Position:** 5% of bankroll. 10% only for Second wave scoring 9+. Half of that
  for Fresh spike, Revival, or an `N` flag.
- **Open positions:** 4 at most.
- **Daily stop:** after a 25% drawdown in a day, stop opening trades until the next day.
- **Reserve:** keep 20% untouched, plus gas.
- **Resize weekly** from the current balance. Never add funds to recover losses.

## Exits

Set all exits at entry. For each open position, check its coin page at least every
2 minutes.

- **Profit ladder:** sell 50% at 2×, another 25% at 4×, trail the last 25% with a
  stop 40% below its high.
- **Stop-loss:** −35%. No exceptions, no averaging down.
- **Signal stop**, exit fully if any of these happens:
  - The hold ratio falls under 0.5.
  - The 1ST IN wallet turns `EXITED`.
  - A large-COST holder turns `EXITED`.
  - Effective holders fall to half of their peak since entry.
  - A new `PRE-TAPE` row appears.
- **Chart stop**, exit fully if either of these happens:
  - VOL 1H ÷ MC under 0.01 for 30 minutes.
  - MC breaks below the holder entry floor.
- **Time stop** (exit if not up 20%): Fresh spike 20 min, Seasoned conviction or
  Second wave 60 min, curve 30 min.
- **Match the wallets' horizon:** if holders have sat through volatility for hours,
  don't exit on the first 15% dip unless a stop fires. A Fresh spike is momentum,
  not conviction.

Sells reach the screen with the same few-second delay as buys, so act on any stop
immediately.

## Execution

- Never chase: if MC has moved more than 15% since the triggering entry, skip.
- Slippage 5–10% on graduated coins, higher on curves but capped by the operator.
- Run a honeypot/sell-tax check before any graduated buy. The terminal does not.
- Copy the contract address from the coin page header. Never trade by ticker alone;
  tickers are cloned.

## Output format

For each candidate evaluated, return:

```json
{
  "token": "0x…",
  "symbol": "NEO",
  "stage": "graduated | curve",
  "pattern": "second_wave | seasoned | fresh_spike | revival | handoff | bags",
  "chart": "alive | dead | unclear",
  "entered": 5, "holding": 4, "effective_holders": 3, "hold_ratio": 0.8,
  "oldest_holder_age_min": 185,
  "score": 8,
  "points": {"1": true, "2": true, "3": true, "...": "…"},
  "skips": [],
  "decision": "enter | watch | skip",
  "size_usd": 5,
  "exits": {"tp": ["2x:50%", "4x:25%", "trail40:25%"], "stop_loss": -0.35,
            "time_stop_min": 60, "entry_floor_mc": 1200000},
  "reason": "one sentence"
}
```

For open positions, on every check return `hold | trim | exit` with the rule that
fired.

## What works against you

- **Paid-to-buy promotions:** a project sends ETH to KOLs who then swap. The ETH
  transfer is filtered, the swap is not. They look like a Fresh spike: many entries
  within minutes, no seasoned holders, hold ratio falling soon after.
- **Dumped allocations:** tokens gifted to insiders never appear as buys, but
  selling them appears as `PRE-TAPE`. Several early in a pair's life means insiders
  are unloading.
- **Bagholders:** look like conviction on the board. Only the chart separates them.
- **Clusters:** several wallets, one operator. Always count effective holders.
- **Losing streaks:** 10+ in a row are normal at these odds. The sizing rules are
  built to survive them; do not loosen them after losses.

## Journal

Log every evaluated candidate and every trade: all fields in the output format,
entry and exit prices, which stop or target closed it, PnL. After 50 closed trades,
report which checklist points and patterns predicted winners, and propose threshold
changes to the operator. Do not change thresholds without approval.