Menu
Price performance report for the project's token — price, volume, liquidity, and context
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
npx skills add https://github.com/aaronjmars/aeon-agent --skill token-reportThe command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
Showing SKILL.md
| type | Skill |
| name | token-report |
| category | crypto |
| description | Price performance report for the project's token — price, volume, liquidity, and context |
| var | |
| tags | ["crypto"] |
| mode | write |
| requires | ["TWITTER_API_KEY","ALCHEMY_API_KEY?","XAI_API_KEY?","BASE_RPC_URL?"] |
| capabilities | ["external_api"] |
${var} — Token contract address. If empty, uses tracked token from MEMORY.md.
This skill reads the token to track from the "Tracked Token" section in memory/MEMORY.md.
## Tracked Token
| Token | Contract | Chain |
|-------|----------|-------|
| AEON | 0xbf8e... | base |
If no tracked token is configured and ${var} is empty, abort silently — no article, no notification.
Read the last 30 days of memory/logs/*.md for prior TOKEN_REPORT_STATE: lines (written in step 8). These are the authoritative source of 1d / 7d / 30d deltas, because a stored price yesterday beats an API window that shifts under you.
A daily token report is only useful if it tells the reader what changed and whether it matters. Snapshots of price, volume, and liquidity are table-stakes; the value is in the verdict. Every section below must either sharpen the verdict or be dropped. No filler, no "N/A", no "no specific context" sentences.
curl -s "https://api.geckoterminal.com/api/v2/networks/base/tokens/CONTRACT_ADDRESS"
curl -s "https://api.geckoterminal.com/api/v2/networks/base/tokens/CONTRACT_ADDRESS/pools?page=1"
# Top pool address from the pools response:
curl -s "https://api.geckoterminal.com/api/v2/networks/base/pools/POOL_ADDRESS/ohlcv/day?aggregate=1&limit=30"
curl -s "https://api.geckoterminal.com/api/v2/networks/base/pools/POOL_ADDRESS/ohlcv/hour?aggregate=1&limit=24"
curl -s "https://api.geckoterminal.com/api/v2/networks/base/pools/POOL_ADDRESS/trades"
If curl fails (sandbox block), retry each URL with WebFetch. If the token endpoint returns no data or 404 after both paths, go to step 9 with TOKEN_REPORT_NO_DATA — do not notify, do not write an article.
curl -s "https://api.dexscreener.com/latest/dex/tokens/CONTRACT_ADDRESS"
Use DexScreener for two things only:
ds=divergent in the sources footer and trust the deeper pool. Do not average.boosted or on trending, add one sentence to the Context section.If DexScreener fails, continue with GT only (ds=fail in footer). Never abort on DS failure.
Low-liquidity pair addendum: the 3% deviation threshold above is calibrated for liquid pairs. On thin pairs it produces false ds=divergent flags from harmless tick noise. If the pair's 24h volume is below $100k, raise the DS deviation threshold to 10% instead of 3% before flagging ds=divergent. The deep-pool-wins rule still applies when the larger threshold is exceeded.
If memory/treasury-wallets.json exists, fetch native-ETH balance for each entry on Base. This optional file declares the project's protocol wallets (treasury, deployer, etc.). If it's absent, skip this step (treasury reporting is off).
WALLETS_FILE="memory/treasury-wallets.json"
[ -f "$WALLETS_FILE" ] && jq -c '.wallets[] | select(.chain=="base")' "$WALLETS_FILE" > /tmp/treasury-wallets.jsonl
For each wallet line, query Base in this fallback order:
Public Base RPC eth_getBalance (primary, keyless — the same endpoint the repo's other Base skills use):
RPC="${BASE_RPC_URL:-https://mainnet.base.org}"
curl -m 10 -s -X POST "$RPC" -H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","method":"eth_getBalance","params":["ADDRESS","latest"],"id":1}'
https://mainnet.base.org is the public, keyless Base JSON-RPC endpoint that wallet-profile, tx-explain, lp-lock, and fund-flow already default to; override with BASE_RPC_URL for an authenticated endpoint (append any key in the URL path, never a -H header — the sandbox blocks env-var expansion in headers). The static -H "Content-Type: application/json" carries no secret, so it is safe. Response is JSON-RPC {"jsonrpc":"2.0","result":"0x<hex_wei>","id":1}. Convert hex → decimal → ÷1e18. If the response has no result, the result is null/non-hex, or it carries an error, mark this wallet eth=fetch_fail and continue.
Why not Etherscan here: the legacy per-chain
api.basescan.org/apihost (BaseScan V1) is deprecated, and the unifiedapi.etherscan.io/v2endpoint gates Base (chainid=8453) behind a paid plan ("Free API access is not supported for this chain"). Neither works keyless, so a plain JSON-RPCeth_getBalanceis the reliable keyless path — matchingwallet-profile/tx-explain.
Alchemy (secondary, only if ALCHEMY_API_KEY is set AND the public RPC failed): use ./secretcurl with the literal {ALCHEMY_API_KEY} placeholder (never $ALCHEMY_API_KEY — the permission layer blocks any command text containing a secret expansion, URL path included; see CLAUDE.md → Network & Secrets):
./secretcurl -m 10 -s -X POST "https://base-mainnet.g.alchemy.com/v2/{ALCHEMY_API_KEY}" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"eth_getBalance","params":["ADDRESS","latest"]}'
Identical JSON-RPC shape and hex → decimal → ÷1e18 conversion as above.
WebFetch fallback (sandbox block on either curl): retry the same POST with WebFetch before declaring fetch_fail.
Compute, per wallet:
eth_balance — decimal ETH, 4 decimals.eth_balance_delta_24h — diff vs yesterday's TREASURY_STATE: log line (omit when prior is missing).Aggregate:
treasury_eth_total — sum across role=treasury wallets only (deployer wallets are operational, not protocol funds).treasury_low_alert — true if treasury_eth_total < 0.01 AND treasury_eth_total > 0 (a zero balance is a config error, not a depletion; do not alarm).If memory/treasury-wallets.json is missing or contains no chain=base entries, set treasury=skip in the sources footer and OMIT the Treasury subsection from the article + notification. Do not invent the section.
From the TOKEN_REPORT_STATE: key=value lines in prior logs, load:
For each:
(~7d) or (~30d) to signal the fallback source.Derived signals:
Z.Z×.volume_in_usd >= 1000 in the trades feed. List the top 3 with direction and size in the "What changed" section if ≥1 exists.Pick exactly one label from the table. Thresholds use today's true deltas from step 3. Evaluate top-to-bottom; the first row whose trigger fully matches wins.
| Label | Trigger |
|---|---|
BREAKOUT | Δprice ≥ +10% AND vol ratio ≥ 2.0 |
BREAKDOWN | Δprice ≤ −10% AND vol ratio ≥ 2.0 |
RALLYING | +3% ≤ Δprice < +10% AND vol ratio ≥ 1.0 |
SLIDING | −10% < Δprice ≤ −3% AND vol ratio ≥ 1.0 |
ACCUMULATING | abs(Δprice) < 3% AND buy/sell ratio ≥ 1.3 AND whale buys ≥ 1 |
DISTRIBUTING | abs(Δprice) < 3% AND buy/sell ratio ≤ 0.7 AND whale sells ≥ 1 |
QUIET | vol ratio < 0.5 AND whale trades = 0 |
CONSOLIDATING | (everything else) |
Do not freelance labels. The verdict drives the lede, the TL;DR, and the notification.
Save to output/articles/token-report-${today}.md:
# $TOKEN — ${today}
**Verdict:** [LABEL] — [≤18 words, citing the 1–2 numbers that drove the label]
## 24h at a glance
| Metric | Now | 24h Δ | vs 7d avg |
|--------|-----|-------|-----------|
| Price | $X.XXXX | ±Y.Y% | — |
| Liquidity | $X.XK | ±Y.Y% | — |
| Volume (24h) | $X.XK | — | Z.Z× |
| Buys / Sells | X / Y | ratio Z.ZZ (yest Z.ZZ) | — |
| Whale trades (≥$1k) | N | — | — |
| FDV | $X.XM | — | — |
## Trend
- **7d:** ±X.X% ([one phrase: rallying, range-bound, rolling over, etc.])
- **30d:** ±X.X% ([one phrase])
## Treasury
[Include ONLY if step 2b returned at least one wallet with a real balance. One row per wallet, role-sorted (treasury first, then deployer, then other). If `treasury_low_alert` is true, lead the section with one sentence naming the floor that was crossed.]
| Wallet | Role | ETH | 24h Δ |
|--------|------|-----|-------|
| 0xf1e9…158e | treasury | X.XXXX | ±Y.YY |
| 0x6797…e3a2 | deployer | X.XXXX | ±Y.YY |
## What changed
[2–4 sentences. Name the specific deltas that matter and the verdict they produced. If whale trades exist, list the top 3 as `buy $1.2K @ $0.0042 · 11:03 UTC`. If liquidity moved >5%, name the pool and the $ amount. Tie every sentence back to the verdict. No filler.]
## Social Pulse
[Only include if the social fetch (Step 6) returns >=2 tweets with >=10 engagement. Lead with a one-line read of the conversation shape, then quote 1-3 tweets with @handle + engagement counts. Otherwise OMIT this section entirely.]
## Context
[Only include when there is a genuine link to known activity: a recent repo release, a broader market regime shift, a boost/trending flag, an on-chain event. If none, OMIT. Never write "no specific context".]
---
*Chart: https://www.geckoterminal.com/base/pools/POOL_ADDRESS*
*Contract: CONTRACT_ADDRESS | Chain: Base*
*Sources: gt=ok · ds=[ok|fail|divergent] · xai=[ok|skip|fail] · treasury=[ok|skip|fetch_fail]*
Section discipline:
−0.05% as 0.0%).Path A - twitterapi.io (primary). Structured tweet objects with exact engagement counts and real permalinks - no fabrication risk, ~700ms. Search the last 24h for the ticker or contract, then keep tweets clearing the engagement bar. Auth via ./secretcurl with the literal {TWITTER_API_KEY} placeholder (never $TWITTER_API_KEY; see CLAUDE.md -> Network & Secrets):
TODAY=$(date -u +%Y-%m-%d)
SINCE=$(date -u -d '1 day ago' +%Y-%m-%d 2>/dev/null || date -u -v-1d +%Y-%m-%d)
Q="$TOKEN_SYMBOL OR CONTRACT_ADDRESS since:$SINCE until:$TODAY"
HTTP=$(./secretcurl -s -o /tmp/tw-token.json -w '%{http_code}' -G "https://api.twitterapi.io/twitter/tweet/advanced_search" \
--data-urlencode "query=$Q" --data-urlencode "queryType=Latest" -H "X-API-Key: {TWITTER_API_KEY}")
echo "twitterapi http=$HTTP bytes=$(wc -c </tmp/tw-token.json)"
# on HTTP 200: keep tweets with >=10 engagement, top 5 by likes
jq -r '[.tweets[] | select((.likeCount + .retweetCount + .replyCount) >= 10)]
| sort_by(-.likeCount) | .[:5][]
| [.author.userName, .likeCount, .retweetCount, .replyCount, .url, .text] | @tsv' /tmp/tw-token.json
Substitute the tracked ticker (cashtag form) for $TOKEN_SYMBOL and the token address for CONTRACT_ADDRESS. On HTTP=200 with >=2 tweets clearing the bar, build the Social Pulse section, exclude obvious bots and generic shill posts, and set xai=ok in the footer.
Path B - xAI Grok x_search (fallback). Only if Path A returned non-2xx, empty, timed out, or TWITTER_API_KEY is unset. Call it in-run with ./secretcurl and the literal {XAI_API_KEY} placeholder (never $XAI_API_KEY; see CLAUDE.md -> Network & Secrets):
./secretcurl -s -X POST "https://api.x.ai/v1/responses" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer {XAI_API_KEY}" \
-d '{
"model": "grok-4-1-fast",
"input": [{"role": "user", "content": "Search X for $TOKEN_SYMBOL or CONTRACT_ADDRESS mentions in the last 24 hours with at least 10 likes. Return up to 5 notable tweets with @handle, engagement counts, and a one-line summary of the claim or vibe. Exclude obvious bots and generic shill posts."}],
"tools": [{"type": "x_search"}]
}'
If neither path returns at least 2 tweets that clear the engagement bar, skip the Social Pulse section and set xai=skip in the footer. On API error on both paths, set xai=fail and skip. If neither TWITTER_API_KEY nor XAI_API_KEY is set, set xai=skip.
Write the compiled report to output/articles/token-report-${today}.md.
Append to memory/logs/${today}.md:
### token-report
- Verdict: [LABEL]
- TOKEN_REPORT_STATE: price=X.XXXX liquidity=XXXX.XX volume_24h=XXXX.XX buys=N sells=N whales=N pool=POOL_ADDRESS
- TREASURY_STATE: treasury_eth_total=X.XXXX wallets=N (one TREASURY_WALLET_STATE line per wallet below)
- TREASURY_WALLET_STATE: addr=0xf1e9…158e role=treasury eth=X.XXXX
- TREASURY_WALLET_STATE: addr=0x6797…e3a2 role=deployer eth=X.XXXX
- 24h: ±X.X% | 7d: ±X.X% | 30d: ±X.X%
- Article: output/articles/token-report-${today}.md
- Sources: gt=ok ds=[ok|fail|divergent] xai=[ok|skip|fail] treasury=[ok|skip|fetch_fail]
The TOKEN_REPORT_STATE: line is a contract — step 3 of the next run parses it with a key=value split. Keep the keys, order, and numeric formats stable. No currency symbols, no thousands separators. The TREASURY_WALLET_STATE: lines (one per fetched wallet) feed step 2b's 24h delta — parse them with the same key=value split, keyed on addr. Omit TREASURY_STATE and TREASURY_WALLET_STATE lines entirely on treasury=skip runs so a wallets.json that disappears later doesn't leave stale balances in the log.
Lead with the verdict, not raw numbers. One short paragraph plus metrics.
*$TOKEN — [LABEL]*
[One sentence citing the driving number(s).]
Price $X.XXXX (±Y.Y% 24h) | Liq $X.XK (±Z.Z%) | Vol $X.XK (W.W× 7d)
Buys/Sells X/Y (ratio Z.ZZ) | Whales: N
Treasury: X.XXXX ETH (±Y.YY 24h)
Chart: https://www.geckoterminal.com/base/pools/POOL_ADDRESS
The Treasury: line is included ONLY when step 2b populated treasury_eth_total > 0. Omit the line entirely on treasury=skip / treasury=fetch_fail runs — silence beats a misleading number.
Skip rules:
TOKEN_REPORT_NO_DATA (step 1 bailout): log only, no notification, no article.QUIET verdict with whales=0 AND abs(Δprice 24h) <1%: send a single-line notification $TOKEN quiet — $X.XXXX flat, vol $X.XK. (no table). This confirms the skill ran without pinging channels with filler on dead days. Exception: if treasury_low_alert is true, override QUIET and send the full notification with a leading *Treasury gas reserve low — X.XXXX ETH on treasury, floor 0.01 ETH.* line. A token going quiet on a day when the agent can no longer pay for gas is the exact regime where the operator needs to see it.Treasury alert (independent of verdict): if treasury_low_alert is true on any run, prepend this line to the notification body — even on QUIET, even on CONSOLIDATING:
⚠️ *Treasury gas reserve low — X.XXXX ETH on treasury (floor 0.01 ETH).*
Public reads (GeckoTerminal, DexScreener, the keyless Base RPC mainnet.base.org) use plain curl; on a real failure (non-2xx / timeout / empty), retry with WebFetch - it accepts the JSON body for the eth_getBalance POST. The auth'd calls (Alchemy fallback, twitterapi.io + api.x.ai Social Pulse) run in-run via ./secretcurl with {ALCHEMY_API_KEY} / {TWITTER_API_KEY} / {XAI_API_KEY} placeholders - no pre-fetch / post-process plumbing.
The Alchemy fallback in step 2b uses $ALCHEMY_API_KEY in the URL path (not in a header), so curl envvar expansion is safe here. If Alchemy is unset, skip silently — the keyless public RPC + WebFetch are enough.
TOKEN_REPORT_NO_DATA, exit silently. No notification about the failure.TOKEN_REPORT_STATE: log line schema — tomorrow's run depends on it.TREASURY_WALLET_STATE: log line schema (one line per fetched wallet, keyed on addr) — step 2b's 24h delta depends on it.treasury=fetch_fail, omit the Treasury subsection + notification line but still write the full token report — the token data is the primary product, treasury is an annotation.role outside {treasury, deployer} appear in the article table sorted last under "other"; only role=treasury wallets count toward treasury_eth_total and the low-balance alert.