Token Pick

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkNone observedNo outbound calls in the source
• ShellNone observedNo commands run in the source
• DependenciesPinnedNothing floats
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass.

Audit 0.4.1 · source sha 0448e87653f9 · 14 Oct 2026Report an issue or request a re-scan

• Publisheraaronjmars GitHub organizationOwner of the repository this skill is read from
• Source folderaaronjmars/aeon/skills/token-pickThe exact folder we audited
• LicenceMITAs the repository states it
• Commit pinned for this verdict0448e87653f9A new audit runs when the source changes
• Repository statusNot archivedPushed today
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: One token recommendation and one prediction market pick - scored, quantified, with a skip branch when signals are weak

---
name: token-pick
description: One token recommendation and one prediction market pick - scored, quantified, with a skip branch when signals are weak
metadata:
  title: Token Pick
  mode: read-only
  category: crypto
  var: ""
  tags:
    - crypto
  requires:
    - COINGECKO_API_KEY?
  capabilities:
    - external_api
    - sends_notifications
---
<!-- autoresearch: variation B — sharper output via signal scoring, edge calculation, conviction tiers, and a skip-day branch -->

> **${var}** — Focus area or thesis (e.g. "AI tokens", "election markets", "contrarian bets"). If empty, scans broadly.

Read `memory/MEMORY.md` for context.
Read the last 7 days of `memory/logs/` and grep for prior `Token Pick` entries — extract the symbols and market questions already picked. **Hard dedup gate**: do not re-pick the same token or the same prediction market unless there is a materially new catalyst that you can name in one sentence.

## Goal

Produce ONE token call and ONE prediction-market call per day, each with a numeric signal/edge score and a conviction tier. If neither qualifies for at least MEDIUM conviction, send a short "no picks today" message rather than forcing a weak pick.

## Steps

### 1. Fetch token data

```bash
# Optional demo key rides ./secretcurl's {COINGECKO_API_KEY} placeholder (never a bare
# $SECRET on the line); without a key the same public endpoint works at a lower rate limit.
CG_HDR=(); [ -n "${COINGECKO_API_KEY:+x}" ] && CG_HDR=(-H "x-cg-demo-api-key: {COINGECKO_API_KEY}")

# Trending coins
./secretcurl -s "${CG_HDR[@]}" "https://api.coingecko.com/api/v3/search/trending"

# Top 250 by market cap with 24h and 7d changes
./secretcurl -s "${CG_HDR[@]}" "https://api.coingecko.com/api/v3/coins/markets?vs_currency=usd&order=market_cap_desc&per_page=250&page=1&sparkline=false&price_change_percentage=24h,7d"

# BTC + ETH 24h/7d for relative-strength benchmark (extract from the markets call above; no extra request needed)

# DEX-side cross-confirmation (no auth, optional but preferred)
curl -s "https://api.dexscreener.com/latest/dex/search?q=trending"
```

If any curl returns empty or errors, retry once with **WebFetch** for the same URL. Track per-source status (`cg=ok|fail`, `dex=ok|fail`) — surfaced in the output footer.

### 2. Fetch prediction markets

```bash
# Top events by 24h volume (events group multi-outcome questions)
curl -s "https://gamma-api.polymarket.com/events?active=true&closed=false&order=volume_24hr&ascending=false&limit=30"

# Newer markets gaining traction
curl -s "https://gamma-api.polymarket.com/markets?closed=false&order=startDate&ascending=false&limit=20"
```

WebFetch fallback on failure. Track `poly=ok|fail`.

### 3. Score every candidate token (0–10 scale)

For each token in the top 250 (and the trending list), compute a signal score:

| Signal | Points |
|---|---|
| 24h price change > 0 | +1 |
| 7d price change > 0 | +1 |
| Both 24h and 7d > +5% | +2 (in addition to above) |
| Appears on CoinGecko trending list | +2 |
| Volume/MarketCap ratio ≥ 0.10 | +2 |
| Volume/MarketCap ratio ≥ 0.20 (replaces above) | +3 |
| Outperforming BOTH BTC and ETH on the 7d | +2 |
| Confirmed on DexScreener trending/gainers (cross-source) | +1 |
| Matches `${var}` thesis when set | +1 |

Drop candidates with market cap < $20M (too pumpable) unless `${var}` explicitly targets micro-caps. Drop any token already picked in the last 7 days (per dedup gate) unless you can name a fresh catalyst.

Pick the highest-scoring token. Use **WebSearch** to surface the most likely catalyst and at least one named risk (regulatory, unlock, narrative-faded, exchange listing, etc.).

### 4. Score prediction markets — edge calculation

For the top ~10 markets by 24h volume that pass the dedup gate (and `${var}` filter when set), do this for each:

1. Read the question and current YES price (`price`/`outcomePrices`).
2. Use **WebSearch** to gather 1–3 recent data points relevant to the resolution.
3. Estimate a **fair YES probability** as a single number (your best calibrated guess, not a range). State the 1–3 inputs you used.
4. Compute `edge = |fair − current_price|` as percentage points.
5. Liquidity gate: require 24h volume ≥ $50k AND market not resolving in < 24h (no last-minute mean-reversion roulette).

Pick the market with the largest edge that clears the gate. If you cannot defend a fair-value estimate within ±10% (insufficient public info), discard and try the next market.

### 5. Conviction tiers

| Tier | Token criterion | Market criterion |
|---|---|---|
| HIGH | signal score ≥ 7 | edge ≥ 10pp |
| MEDIUM | signal score 4–6 | edge 5–10pp |
| SKIP | signal score < 4 | edge < 5pp |

**Skip-day branch**: if BOTH the chosen token and the chosen market land in SKIP, do not synthesize a pick. Send the skip message (step 6b) and log accordingly. This is a feature — forcing low-conviction picks degrades the signal of the whole feed.

### 6a. Notification — normal day (under 4000 chars)

Send via `./notify`:

```
*Daily Pick — ${today}*

*Token: SYMBOL*  [HIGH | MEDIUM]  signal X/10
Price: $X.XX (±X.X% 24h / ±X.X% 7d) | mcap $XB | vol $XM (vol/mcap X.XX)
Score breakdown: [trending+2, vol/mcap+3, RS vs BTC/ETH+2, narrative+1] = 8/10
Catalyst: [one sentence — what's driving this right now, named source/event]
Risk: [one sentence — concrete risk, not generic "could go down"]
Vs recent picks: [first time / repeat with new catalyst: ...]

*Market: "Question?"*  [HIGH | MEDIUM]  edge Xpp
Current: YES X¢ / NO Y¢ | 24h vol $Xm | resolves: DATE
Fair YES: ~Y% (inputs: [src1], [src2], [src3])
Thesis: [one sentence — why the market is wrong, action implied]
Risk: [one sentence — what could make your fair-value estimate wrong]

sources: cg=ok|fail, dex=ok|fail, poly=ok|fail
not financial advice — pattern-matching only
```

If only one of the two pick types qualifies, send just that one section (omit the other entirely — do not include a HIGH and a SKIP in the same message).

### 6b. Notification —

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

git clone https://github.com/aaronjmars/aeon
git -C aeon checkout 0448e87653f9

Works in

1. 2026-10-14
   ✓SAFE B 89 / 100

   first audit · source 0448e87653f9

Questions about Token Pick

• github.com/aaronjmars/aeon/tree/main/skills/token-pick — the audited source at commit 0448e87653f9 and every finding aboveObserved · 14 Oct 2026

Spotted something wrong? Report a correction · Methodology