How a closed swap becomes a graded post-mortem
Four steps to set up, four passes over every trade after that. You do the first part once and never touch it again.
Most trading journals die the same way. They ask you to export a CSV, tidy the columns, and write down what you were thinking. That works for about two weeks. The trades you most need to review are the ones you least want to type out, so those are the first to go missing, and a journal with the bad trades missing is worse than no journal at all.
Dossier reads the chain instead. Your wallet already recorded everything, in public, at the time, without editorialising. The work is turning that record into something you can learn from.
Setting it up
01 · Connect. Connect wallet, or just email
One free, off-chain message signature proves the wallet is yours. Never a transaction, never a token approval. From there, everything is public-ledger reads.
02 · Auto-import. Every closed trade, captured
Buys, sells, partial exits. Each closed position becomes a journal entry. New trades land in seconds. The graded post-mortem typically follows in under a minute.
03 · Analyze. AI writes the post-mortem
Chart reconstruction, token risk, entry timing, exit timing, comparison to your recent trades and your 90-day behavior profile. One paragraph, twelve to forty seconds of reading.
04 · Improve. Patterns become a checklist
A one-page Trader Report Card, recomputed with every graded trade. Your worst hour, worst category, single biggest leak. The list shrinks every week, or it doesn’t.
What happens to every closed trade
Once a position closes, four passes run over it. New trades are detected within seconds, and the graded post-mortem typically follows in under a minute.
Pass 01. Reconstruct the chart you actually traded
1-minute candles, or 15-second bars on scalps, from 15 minutes before entry to 60 minutes after exit. The exact context you saw, on an interactive chart.
Pass 02. Pull the token’s full provenance
Bundler-launch flag, launch time, deployer wallet, 1–10 risk score. Was this token a coin flip or a coordinated trap?
Pass 03. Run the AI coach over the full picture
Reads the chart, the metadata, your entry timing, your exit timing, your recent trades, and a rolling 90-day profile of how you actually trade. Writes one paragraph. No fluff. No emoji.
Pass 04. Letter grade. A through F.
No 0-100 score, no “trade quality index,” no five-axis radar chart. Plain school grades, because you already know what they mean.
What a position is
Dossier groups buys, sells and partial exits into the position they belong to, rather than logging raw swaps. Scaling into a token across four buys and out across three sells produces one graded entry, not seven unrelated rows.
That grouping is what makes a grade meaningful. Judging a single partial exit in isolation tells you nothing about whether the trade was any good.
What it never asks of you
- No CSV export, no spreadsheet, no manual reconciliation.
- No transaction approval and no token spend permission, ever.
- No private key or seed phrase. Neither is requested or stored.
- No writing up your own trades. The record is already on chain; the analysis is the part you were missing.
The full detail of what a wallet connection can and cannot do is on the security page.
What you get out of it
Each closed trade ends up with a letter grade, a reconstructed chart you can actually look at, tagged mistakes drawn from a fixed taxonomy of ten, and one paragraph explaining what the data says set the trade up.
Repeat offences get flagged from their second occurrence and carry a running count, which is the point. One bad trade is noise. The same bad trade fourteen times is a leak with a name, and the grade is how you notice it before the count gets that high.