analyst

name	analyst
description	Deep fundamental analysis and scoring of discovered tokens
triggers	["analyze token","evaluate token","deep dive","score this token","what do you think about","analysis"]

Note: This skill requires deep reasoning. Gather all data (token-metrics, check-contract, holder-distribution outputs, current portfolio, relevant memory patterns) before executing.

Analyst Skill

Purpose

Evaluate every token discovery on fundamentals, tokenomics, social sentiment, narrative fit, and timing. Produce a scored assessment.

When to Use

• After discovery skill finds new tokens
• When user asks to analyze a specific token
• When reassessing an existing position

Process

Step 1: Gather Deep Data

Get full token metrics:

node scripts/token-metrics.js --address <TOKEN_ADDRESS> --chain <CHAIN>

Check contract safety:

node scripts/check-contract.js --address <TOKEN_ADDRESS> --chain <CHAIN>

Two-source confirmation (PR 4.2): the executor enforces, at signing time, that BOTH DEXScreener and Birdeye see the token and agree on price within 2%. Tokens that fail this gate are auto-quarantined with a Telegram alert. You don't need to verify two-source agreement during analysis — the gate fires regardless — but if you find a token that DEXScreener has and Birdeye doesn't, expect quarantine. The 2-source check exists because single-source tokens (typically newly-listed, obscure venue, or post-list-but-pre-Birdeye-indexing) carry concentrated rug risk.

Recovered safety rejections (e.g. GoPlus "Not fungible SPL token address"): when check-contract.js returns a structural reject like not fungible, not a token, or no holder data, this is a cached avoid decision — the token shouldn't be analyzed further, but the pipeline didn't crash. Cache the verdict via cache-analysis --json '{"verdict":"avoid","reason":"<reason>",...}' and log to add-research-log with "status":"warning" (NOT "error"). Reserve status:"error" for unrecovered crashes (network timeouts, JSON parse failures with no usable output). Observer treats error as a silent crash and files an issue per occurrence.

Check holder distribution:

node scripts/holder-distribution.js --address <TOKEN_ADDRESS> --chain <CHAIN> --propose

Step 2: Score Across 6 Dimensions

Contract Safety (Weight: 25%)

Signal	Points
Verified + renounced	90-100
Verified, not renounced but multisig	70-85
Verified, single owner	40-60
Unverified	0-20
Mint function present	-20
Proxy/upgradeable	-15
Audit by known firm	+15

Tokenomics (Weight: 20%)

Signal	Points
Top 10 wallets < 30%	80-100
Top 10 wallets 30-50%	40-70
Top 10 wallets > 50%	0-30
Team allocation < 10% with lock	+15
Deflationary mechanics	+10
Clear vesting schedule	+10

Liquidity (Weight: 20%)

Signal	Points
LP locked > 6 months	85-100
LP locked 1-6 months	60-80
LP burned	90-100
LP not locked	0-30
Liquidity:MCap ratio > 10%	+15
Liquidity > $100k	+10

Social & Community (Weight: 15%)

Signal	Points
Organic community, active devs	80-100
Good community, quiet devs	50-70
Mostly bots / paid shills	0-30
Influencer mentions (organic)	+10
Influencer mentions (paid only)	-10

Narrative Fit (Weight: 10%)

26 narratives tracked (see scripts/narrative-config.js for full list): ai_infra, ai_agents, defi, restaking, lst, yield, payfi, rwa, prediction, l2, zk, modular, intents, depin, memecoin, socialfi, gaming, nft_infra, btc_eco, btc_l2, privacy, telegram, consumer, desci, degov, energy.

Signal	Points
Leading token in hot narrative	90-100
Solid token in active narrative	60-80
Narrative cooling down	30-50
No clear narrative	0-20

Apply the tier affinity boost from the "Narrative score modifier" subsection of Step 4 on top of the base score (cap at 100).

Timing (Weight: 10%)

Signal	Points
Early (< 24h, metrics growing)	85-100
Mid (1-7d, momentum building)	60-80
Late (> 7d, price already 5x+)	20-40
Volume trend increasing	+10
At support level	+10

Step 3: Calculate Weighted Score

overall = (contract * 0.25) + (tokenomics * 0.20) + (liquidity * 0.20) +
          (social * 0.15) + (narrative * 0.10) + (timing * 0.10)

Step 4: Assign Tier

Based on token characteristics AND narrative context, assign the appropriate portfolio tier.

Base tier check

Criteria	Tier
Token address matches any entry in baseTierTokens from get-chain-config --chain <CHAIN> AND base is in the chain's tiersEnabled	base

Base tier is a closed set. ONLY the tokens listed in baseTierTokens from get-chain-config can be base tier. No other token — regardless of liquidity, age, market cap, or how underweight the base allocation is — may be classified as base. If a token is not in baseTierTokens, or if base is not in the chain's tiersEnabled, proceed to the narrative-aware assignment below.

Narrative-aware tier assignment

Each narrative has a tier affinity (defined in scripts/narrative-config.js). Use this decision tree:

Is tierAffinity "strong_moonshot"?
  → Always moonshot. Volatility too high for conviction sizing.
    Narratives: memecoin, socialfi, telegram, desci, energy, degov

Is tierAffinity "lean_moonshot"?
  → Default moonshot. Only conviction if EXCEPTIONAL:
    age >30 days AND liquidity >$500k
    Narratives: ai_agents, gaming, btc_eco, btc_l2, nft_infra, intents, consumer

Does token meet conviction criteria?
  (age >7d, liquidity >$100k, mcap >$1M, verified + renounced/multisig)

  YES + affinity "strong_conviction" or "lean_conviction"?
    → conviction
    Narratives: defi, lst, restaking, rwa, l2, zk, modular,
                ai_infra, yield, depin, payfi, prediction, privacy

  YES + affinity "lean_moonshot"?
    → moonshot (unless exceptional as above)

  NO → moonshot

Narrative score modifier

When narrative is hot or warming, apply a boost to the narrative fit dimension score:

Tier Affinity	Narrative Score Boost
strong_conviction	+15 when hot/warming
lean_conviction	+10 when hot/warming
lean_moonshot	+5 when hot
strong_moonshot	+0 (no boost)

The tier determines position limits, stop-loss levels, and slippage tolerance.

Step 5: Recommendation

Score	Recommendation
71-100	strong_buy — high conviction, act soon
51-70	buy — solid opportunity
31-50	watch — interesting, wait for better entry
0-30	avoid — too many red flags

Step 6: Output Format

{
  "tokenAddress": "string",
  "chain": "string",
  "symbol": "string",
  "scores": {
    "contract": 0,
    "tokenomics": 0,
    "liquidity": 0,
    "social": 0,
    "narrative": "string",
    "narrativeScore": 0,
    "timing": 0,
    "overall": 0
  },
  "strengths": ["string"],
  "weaknesses": ["string"],
  "recommendation": "strong_buy | buy | watch | avoid",
  "tier": "base | conviction | moonshot",
  "suggestedEntry": 0,
  "suggestedSize": "small | medium | large",
  "reasoning": "2-3 sentence summary"
}

Step 7: Log & Handoff

• Write to memory/YYYY-MM-DD.md with [ANALYSIS] tag
• Check MEMORY.md for similar past analyses and their outcomes
• If recommendation is buy or strong_buy → proceed to risk skill
• If watch → add to watchlist via database:

  node scripts/db-query.js add-to-watchlist --json '{"symbol":"TOKEN","address":"0x...","chain":"<CHAIN>","reason":"...","target_entry":0.001}'
  

• If avoid → cache the result and end:

  node scripts/db-query.js cache-analysis --json '{"address":"<TOKEN_ADDRESS>","chain":"<CHAIN>","symbol":"<SYMBOL>","analysis_score":<SCORE>,"verdict":"avoid","reasoning":"<REASON>"}'
  

Rules

• NEVER let excitement override analysis
• If something looks too good, look harder for the catch
• Acknowledge uncertainty — if you can't verify something, adjust score down
• Compare against alternatives in the same narrative
• Always suggest a specific entry price, not just "buy"

Promotion

If an analysis pattern (e.g., a scoring heuristic, token archetype, or recurring outcome correlation between metrics and post-trade results) recurs 3+ times across daily logs, promote it to MEMORY.md using the template in AGENTS.md § MEMORY.md Updates.