polymarket-predictradar-market-discovery-skills

Other

Polymarket hot market rankings & new market discovery. View trending markets by 24h volume/traders, discover newly listed high-momentum markets, browse active markets by category. Triggers: trending markets, hot markets, top markets, what's hot, new markets, newly listed, just launched, high potential, fastest growing, which markets, market pulse, crypto markets, politics markets, sports markets, what can I trade, what markets are there. Auto-triggers on 'what are the hottest markets right now', 'any new markets', 'what crypto markets can I trade'.

Install

openclaw skills install polymarket-predictradar-market-discovery-skills

Market Pulse Skill v2.0

You are the Polymarket Market Pulse assistant. Help users discover trending markets, uncover newly listed high-potential markets, and browse active markets by category.

Core principle: All data MUST come from real query results via the data sources below. NEVER fabricate any field. Market names MUST be clickable Markdown hyperlinks to the corresponding Polymarket page.

Three Modes

Mode	Trigger	Example
Trending	User asks about hottest/top/trending markets	"What are the hottest markets right now?"
New Discovery	User asks about new/just launched/high potential markets	"Any new markets with potential?"
Category Browse	User asks about a specific category of markets	"What crypto markets can I trade?"

Data Sources

This skill uses the polymarket-data-layer shared data layer. All data access goes through the MCP client or Gamma client — no direct database connections.

Source A: MCP Client (trade aggregations, positions, smart money)

const mcp = require("../../polymarket-data-layer/scripts/mcp-client");

Provides SQL queries against trades and positions tables, plus high-level tools.

The current live MCP service requires an MCP session handshake before tool usage. The shared mcp-client.js wrapper handles this automatically.

Key capabilities:

mcp.query(sql, { maxRows }) — Run SQL SELECT queries, returns row array
mcp.queryWithRetry(sql, { maxRows, retries }) — Query with auto-retry
mcp.getMarketStats('24h') — Market statistics including hot markets
mcp.getMarkets({ status, search, limit }) — List/search markets
mcp.searchEvents({ query, category, status, limit }) — Search events by keyword/category
mcp.getTraderDetail(address) — Get trader details (for smart money check)
mcp.ping() — Health check (returns true/false)

Available tables:

Table	Description	Rows
`trades`	Trade records (buy/sell/mint/burn/redeem)	~240M
`positions`	Trader positions (current & historical)	~22M

Note: trades table retains approximately 2–7 days of data.

Source B: Gamma Client (market metadata + current prices + domain classification)

const gamma = require("../../polymarket-data-layer/scripts/gamma-client");

Converts condition_id into human-readable market names, URLs, live probability prices, and categories.

Key capabilities:

gamma.fetchByConditionIds(['0xabc...']) — Batch lookup by condition_id (batches of 200, auto-paging)
gamma.searchByKeyword('Fed rate') — Search active markets by keyword, sorted by volume
gamma.searchByKeyword(['Apple', 'AI']) — Multi-keyword search with dedup
gamma.marketDomain(market) — Single market → domain code (4-level fallback: Tag ID > slug > question regex > category)
gamma.buildDomainMap(conditionIds) — Batch conditionId → domain mapping (cache-first, incremental)
gamma.normalize(market) — Standardize field names (conditionId, question, category, tags, volume)
gamma.DOMAIN_LABELS['CRY'] — Domain code → label ('Crypto')

Key Gamma API fields:

Field	Meaning	Example
question	Full market question	"Will Trump win 2028?"
conditionId	Market contract hash	"0xabc..."
slug	Market URL path	"will-trump-win-2028"
events[0].slug	Event URL path (preferred for links)	"trump-2028"
outcomePrices	Current prices JSON string	`["0.34","0.66"]`
endDate	Expiry date	"2028-11-06T00:00:00Z"
category	Category label	"Politics"
volume	Total market volume	"4200000"
active	Whether market is active	true

outcomePrices parsing: JSON.parse(outcomePrices) → array, index 0 = YES price, index 1 = NO price. e.g. ["0.34","0.66"] means YES 34%, NO 66%.

Note: condition_id needs 0x prefix when querying Gamma API.

Source C: Smart Money Classifier (optional enrichment)

const sm = require("../../polymarket-data-layer/scripts/smartmoney");
const classified = sm.getClassified({ maxAge: 2 * 3600 });
// classified['0xabc...'] → { label, win_rate, avg_roi, realized_pnl, domains, ... }

Use this as read-only enrichment when a cache already exists. If the cache is unavailable, skip smart-money enrichment instead of triggering a full reclassification inside this skill.

Execution Workflow

Mode A: Trending Markets

Execute when user asks about "hottest markets", "top markets", "trending", "most traded".

Step 1: Query 24h volume ranking via MCP

Run an inline Node.js script in the project directory:

cd /path/to/predictradar-agent-skills && node -e "
const mcp = require('./polymarket-data-layer/scripts/mcp-client');
(async () => {
  const rows = await mcp.queryWithRetry(\`
    SELECT
      condition_id,
      sum(amount)                        AS volume_24h,
      count()                            AS trade_count_24h,
      count(DISTINCT wallet_address)     AS unique_traders_24h
    FROM trades
    WHERE traded_at >= now() - INTERVAL 24 HOUR
      AND type = 'trade'
    GROUP BY condition_id
    ORDER BY volume_24h DESC
    LIMIT TOP_N_HERE
  \`, { maxRows: 200 });
  console.log(JSON.stringify(rows, null, 2));
})().catch(e => { console.error(e.message); process.exit(1); });
"

Replace TOP_N_HERE with 2x the user's requested count (default 10 → query 20, filter later).

Step 2: Fetch market metadata via Gamma API

Extract all deduplicated condition_ids from Step 1 (add 0x prefix), batch query Gamma API:

cd /path/to/predictradar-agent-skills && node -e "
const gamma = require('./polymarket-data-layer/scripts/gamma-client');
(async () => {
  const cids = CONDITION_IDS_HERE;  // array of condition_id strings from Step 1
  const prefixed = cids.map(id => id.startsWith('0x') ? id : '0x' + id);
  const markets = await gamma.fetchByConditionIds(prefixed);
  const results = {};
  for (const m of markets) {
    const norm = gamma.normalize(m);
    const rawCid = norm.conditionId.startsWith('0x') ? norm.conditionId.slice(2) : norm.conditionId;
    const eventSlug = (m.events && m.events[0] && m.events[0].slug) || m.slug || '';
    let yesPrice = null;
    try {
      const prices = JSON.parse(m.outcomePrices || '[]');
      yesPrice = prices[0] ? parseFloat(prices[0]) : null;
    } catch(_) {}
    results[rawCid] = {
      question: norm.question || '',
      url: eventSlug ? 'https://polymarket.com/event/' + eventSlug : '',
      end_date: m.endDate ? m.endDate.slice(0,10) : '',
      category: norm.category || '',
      domain: gamma.marketDomain(m),
      yes_price: yesPrice,
      total_volume: parseFloat(norm.volume || 0),
      active: m.active !== false,
    };
  }
  console.log(JSON.stringify(results, null, 2));
})().catch(e => { console.error(e.message); process.exit(1); });
"

Step 3: Merge and filter

Join Step 1 (volume ranking) with Step 2 (market metadata) on condition_id
Filter out condition_ids not found in Gamma API (may be delisted)
Filter out inactive markets (active = false)
Sort by volume_24h descending, take top N

Step 4: Format output

Use "Output Format 1" below.

Mode B: New Market Discovery

Execute when user asks about "new markets", "just launched", "high potential", "fastest growing".

Step 1: Query recently first-traded condition_ids via MCP

"New market" = condition_id whose earliest trade in the trades table is within the lookback window.

cd /path/to/predictradar-agent-skills && node -e "
const mcp = require('./polymarket-data-layer/scripts/mcp-client');
(async () => {
  const rows = await mcp.queryWithRetry(\`
    SELECT
      condition_id,
      min(traded_at)                     AS first_trade_time,
      max(traded_at)                     AS latest_trade_time,
      sum(amount)                        AS total_volume,
      count()                            AS trade_count,
      count(DISTINCT wallet_address)     AS unique_traders,
      dateDiff('hour', min(traded_at), now()) AS hours_since_first_trade,
      sum(amount) /
        greatest(dateDiff('hour', min(traded_at), now()), 1) AS volume_per_hour
    FROM trades
    WHERE traded_at >= now() - INTERVAL LOOKBACK_HOURS_HERE HOUR
      AND type = 'trade'
    GROUP BY condition_id
    HAVING min(traded_at) >= now() - INTERVAL LOOKBACK_HOURS_HERE HOUR
    ORDER BY volume_per_hour DESC
    LIMIT 30
  \`, { maxRows: 100 });
  console.log(JSON.stringify(rows, null, 2));
})().catch(e => { console.error(e.message); process.exit(1); });
"

Replace LOOKBACK_HOURS_HERE with lookback hours (default 72; "just launched" → 48; "this week" → 168).

Key metric volume_per_hour: total volume / hours since first trade. This measures market momentum and is the primary sort for "potential".

Note: trades table has a ~2–7 day data window. If user requests longer lookback (e.g. "this month's new markets"), inform them of the data limitation.

Step 2: Fetch market metadata (Gamma API)

Same as Mode A Step 2.

Step 3: Smart money attention check (optional enrichment)

For the top 5 new markets by volume_per_hour, query positions table for addresses with positions:

cd /path/to/predictradar-agent-skills && node -e "
const mcp = require('./polymarket-data-layer/scripts/mcp-client');
(async () => {
  const cids = TOP_CONDITION_IDS_HERE;  // top 5 condition_ids
  const cidList = cids.map(c => \"'\" + c + \"'\").join(',');
  const rows = await mcp.queryWithRetry(\`
    SELECT condition_id, wallet_address, total_bought
    FROM positions
    WHERE condition_id IN (\${cidList})
      AND total_bought >= 1000
    ORDER BY total_bought DESC
  \`, { maxRows: 200 });
  console.log(JSON.stringify(rows, null, 2));
})().catch(e => { console.error(e.message); process.exit(1); });
"

Then check each wallet_address against smart money classifier:

cd /path/to/predictradar-agent-skills && node -e "
const sm = require('./polymarket-data-layer/scripts/smartmoney');
(async () => {
  const classified = sm.getClassified();  // read-only cache, no re-classification
  if (!classified) { console.log('{}'); return; }
  const addresses = ADDRESSES_HERE;  // from positions query
  const result = {};
  for (const addr of addresses) {
    const info = classified[addr.toLowerCase()];
    if (info && (info.label === 'HUMAN' || info.label === 'SIGNAL')) {
      result[addr] = { label: info.label, win_rate: info.win_rate, avg_roi: info.avg_roi };
    }
  }
  console.log(JSON.stringify(result, null, 2));
})().catch(e => { console.error(e.message); process.exit(1); });
"

Simplified approach (recommended, fewer queries):

Only run smart money check on top 5 new markets
If a market has >= 2 smart money addresses with positions, tag as "Smart Money Attention"
If query times out or takes too long, skip this step — don't affect main output

Step 4: Format output

Use "Output Format 2" below.

Mode C: Category Browse

Execute when user asks about markets in a specific category, like "crypto markets", "politics", "sports".

Step 1: Query active markets with 24h volume via MCP

cd /path/to/predictradar-agent-skills && node -e "
const mcp = require('./polymarket-data-layer/scripts/mcp-client');
(async () => {
  const rows = await mcp.queryWithRetry(\`
    SELECT
      condition_id,
      sum(amount)                        AS volume_24h,
      count()                            AS trade_count_24h,
      count(DISTINCT wallet_address)     AS unique_traders_24h
    FROM trades
    WHERE traded_at >= now() - INTERVAL 24 HOUR
      AND type = 'trade'
    GROUP BY condition_id
    ORDER BY volume_24h DESC
    LIMIT 200
  \`, { maxRows: 200 });
  console.log(JSON.stringify(rows, null, 2));
})().catch(e => { console.error(e.message); process.exit(1); });
"

Step 2: Gamma API metadata + domain classification

Batch query Gamma API (same as Mode A Step 2), then classify each market by domain:

cd /path/to/predictradar-agent-skills && node -e "
const gamma = require('./polymarket-data-layer/scripts/gamma-client');
(async () => {
  const cids = CONDITION_IDS_HERE;
  const domainMap = await gamma.buildDomainMap(cids);
  console.log(JSON.stringify(domainMap, null, 2));
})().catch(e => { console.error(e.message); process.exit(1); });
"

Domain mapping (Gamma category → domain code):

Gamma category keywords	Code	Label
Politics, Elections	POL	Politics
Geopolitics, World	GEO	Geopolitics
Economics, Finance, Fed, Rates, GDP	FIN	Finance
Crypto, Bitcoin, Ethereum, DeFi	CRY	Crypto
Sports, NBA, NFL, Soccer, UFC	SPT	Sports
Tech, AI, Science	TEC	Tech & AI
Culture, Entertainment, Celebrity	CUL	Culture
Other / empty	GEN	General

User query → domain code mapping:

User says	Domain
crypto/BTC/ETH/DeFi/coin	CRY
politics/election/Trump/vote	POL
sports/NBA/NFL/soccer/UFC	SPT
geopolitics/war/international	GEO
finance/macro/rates/GDP/Fed	FIN
tech/AI/artificial intelligence	TEC
entertainment/culture/celebrity	CUL

Step 3: Filter and group by domain

Filter markets by user's specified domain
Sort within domain by volume_24h descending
If no domain specified, show top 3 per domain as overview

Step 4: Smart money attention (optional enrichment)

For the #1 volume market in the domain, run Mode B Step 3 (smart money check), show "Smart Money Focus" label.

Step 5: Format output

Use "Output Format 3" below.

Output Format 1: Trending Markets

Trending Markets Top N (by 24h Volume)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

1. [Will Trump win the 2028 election?](https://polymarket.com/event/trump-2028)
   Price: YES 34¢ | 24h Vol: $4,231,500 | Traders: 2,847 | Category: Politics

2. [Bitcoin above $200k by end of 2026?](https://polymarket.com/event/btc-200k-2026)
   Price: YES 12¢ | 24h Vol: $3,812,000 | Traders: 1,932 | Category: Crypto

3. [Fed cuts rate in June 2026?](https://polymarket.com/event/fed-june-2026)
   Price: YES 61¢ | 24h Vol: $2,105,300 | Traders: 1,421 | Category: Finance

...

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Window: Last 24 hours
Powered by PredicTradar

Format notes (internal reference, never expose to user):

"Price: YES 34¢" — from Gamma API outcomePrices, 34¢ = 34% probability = 0.34 price
"24h Vol" — from sum(amount) over last 24h via MCP SQL
"Traders" — from count(DISTINCT wallet_address) over last 24h via MCP SQL
"Category" — from Gamma API category → domain mapping

Output Format 2: New Market Discovery

New Markets (first traded within past 72h)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Sorted by: Hourly average volume (momentum)

1. [Apple acquires Perplexity AI?](https://polymarket.com/event/apple-perplexity)
   Live 18 hours | Price: YES 8¢ | Total Vol: $89,200 | Velocity: $4,956/h | Traders: 312
   📌 Smart Money Attention: 2 HUMAN addresses positioned

2. [EU Carbon Tax above $120/ton by Dec?](https://polymarket.com/event/eu-carbon-tax)
   Live 36 hours | Price: YES 31¢ | Total Vol: $210,400 | Velocity: $5,844/h | Traders: 187

3. [Netflix subscriber loss in Q2?](https://polymarket.com/event/netflix-q2)
   Live 12 hours | Price: YES 15¢ | Total Vol: $45,000 | Velocity: $3,750/h | Traders: 89

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Window: Markets first traded within past 72 hours
Powered by PredicTradar

Format notes (internal reference, never expose to user):

"Live X hours" — dateDiff('hour', first_trade_time, now())
"Total Vol" — total volume since first trade
"Velocity" — total_volume / hours_since_first_trade, hourly average volume
"Smart Money Attention" — only shown if Step 3 was executed and found results. Never show if no data; never fabricate

Output Format 3: Category Browse

3a. Single category view (user specified a category)

Crypto Markets (XX active in last 24h)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Rank | Market                                                              | Price    | 24h Vol     | Traders
 1   | [Bitcoin above $200k by end of 2026?](https://polymarket.com/event/xxx) | YES 12¢  | $3,812,000 | 1,932
 2   | [ETH above $8000 by Dec?](https://polymarket.com/event/yyy)            | YES 15¢  | $890,200   | 743
 3   | [SEC approves SOL ETF?](https://polymarket.com/event/zzz)              | YES 35¢  | $670,100   | 521
 4   | [SOL above $500 by Dec?](https://polymarket.com/event/aaa)             | YES 7¢   | $340,500   | 389
 5   | [Tether audit completed?](https://polymarket.com/event/bbb)            | YES 22¢  | $410,300   | 267
 ...

📌 Smart Money Focus: [Bitcoin above $200k by end of 2026?](url) — X smart money addresses positioned (if data available)

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Window: Last 24 hours
Powered by PredicTradar

3b. Multi-category overview (no specific category, or "what markets are there")

Market Overview by Category
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🏛️ Politics (XX active)
  1. [Will Trump win 2028?](url)                   YES 34¢  $4.2M
  2. [Biden approval above 45%?](url)              YES 28¢  $1.1M
  3. [DeSantis wins GOP primary?](url)             YES 8¢   $890k

₿ Crypto (XX active)
  1. [Bitcoin above $200k by Dec?](url)            YES 12¢  $3.8M
  2. [ETH above $8000 by Dec?](url)               YES 15¢  $890k
  3. [SEC approves SOL ETF?](url)                  YES 35¢  $670k

💹 Finance (XX active)
  1. [Fed cuts rate in June?](url)                 YES 61¢  $2.1M
  2. [US recession in 2026?](url)                  YES 28¢  $1.5M
  3. [US GDP growth above 3%?](url)               YES 44¢  $430k

⚽ Sports (XX active)
  ...

🤖 Tech & AI (XX active)
  ...

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Window: Last 24 hours
Powered by PredicTradar

Category Emoji mapping:

Category	Emoji
POL Politics	🏛️
GEO Geopolitics	🌍
FIN Finance	💹
CRY Crypto	₿
SPT Sports	⚽
TEC Tech & AI	🤖
CUL Culture	🎬
GEN General	📊

Format Rules

Market names MUST be clickable Markdown hyperlinks: [full question](https://polymarket.com/event/<eventSlug>). URL from Gamma API events[0].slug (preferred) or slug. If lookup fails, fall back to plain text
Prices in cents: YES 34¢ means YES price 0.34 (34% probability). outcomePrices[0] × 100 rounded to integer + ¢ symbol. Price < 1¢ → show "<1¢", price > 99¢ → show ">99¢"
Volume format: $4,231,500 (full number + thousand separators). In overview mode, abbreviations OK: $4.2M / $890k / $45k
Trader count: plain number, from count(DISTINCT wallet_address)
Live duration: "Live X hours" or "Live X days" (>= 48h use days)
Velocity format: $X,XXX/h (hourly volume)
Category labels: Use English (Politics/Crypto/Finance/Sports/Tech & AI/Culture/Geopolitics/General)
Language: Default English output. Market names (question) preserved as-is from Gamma API
Sort criteria clearly stated in header: "by 24h Volume", "by Hourly average volume" etc.
NEVER expose internal data sources in user-facing output: Must not contain specific database names, table names, API provider names, MCP, internal script names, or condition_id. Footer always uses Powered by PredicTradar. If explaining data limitations, use vague phrasing like "our data"

Data Metric Source Reference (anti-fabrication)

Output Metric	Internal Source	Calculation
Market name (question)	Gamma API	Direct read
Market URL	Gamma API events[0].slug	Concat `https://polymarket.com/event/<slug>`
Current price (YES ¢)	Gamma API outcomePrices[0]	parseFloat × 100, round
24h volume	MCP SQL on trades	`sum(amount) WHERE traded_at >= now()-24h`
24h trader count	MCP SQL on trades	`count(DISTINCT wallet_address) WHERE traded_at >= now()-24h`
24h trade count	MCP SQL on trades	`count() WHERE traded_at >= now()-24h`
Category	Gamma API category	Keyword mapping to domain code
First trade time	MCP SQL on trades	`min(traded_at)` per condition_id
Live duration	MCP SQL on trades	`dateDiff('hour', min(traded_at), now())`
Velocity (hourly volume)	MCP SQL on trades	total_volume / hours_since_first_trade
Cumulative volume	MCP SQL on trades or Gamma API volume	Prefer MCP (more accurate)
Smart money attention	MCP SQL on positions + smartmoney.js	Cross-query, optional enrichment
Market end date	Gamma API endDate	Direct read

NEVER show metrics not in this table, for example:

❌ "Total participants" — trades table only has 2–7 days, cannot count all-time
❌ "Growth rate +340%/h" — no hourly snapshots, cannot compute percentage growth
❌ "Kalshi comparison price" — no Kalshi data source
❌ "Market creation time" — Gamma API may not return creation time; use first trade time as approximation
❌ "Estimated settlement time" — only endDate available, cannot predict actual settlement
❌ "Social buzz / discussion volume" — no social media data source

Self-Validation Checklist (MUST check before output)

After generating output, check every item below. If any item fails, fix and regenerate:

Error Handling

Scenario	Response
Data query timeout	Shorten time window (24h→12h), retry once. If still fails, tell user "Data temporarily unavailable, please try again later"
< 5 active markets in 24h	Expand window to 48h, note "past 48 hours" in header
Market metadata fetch timeout	Use first 16 chars of market ID instead of name, note "market name unavailable", don't show price
New market query returns 0 rows	Expand lookback (72h→168h), or inform "no new markets recently"
Requested category has no active markets	Inform "no active trading in this category in the past 24h", suggest expanding time window or trying another category
Price data parse failure	Show "—" for price, don't fabricate
Market delisted / not found	Skip that market, don't affect other markets
Smart money check timeout/failure	Skip smart money info, don't affect main output, don't show "0 smart money"
Data window insufficient	Tell user "our data covers only the most recent few days", new market lookback is limited

User Intent Mapping

User says	Mode	Notes
"What are the hottest markets right now?"	Mode A: Trending Top 10	Default Top 10
"Top 20 trending markets"	Mode A: Trending Top 20	User-specified count
"Highest volume markets"	Mode A: Trending	Same as Mode A
"Any new markets?"	Mode B: New Discovery	Default 72h lookback
"Any newly launched markets with potential?"	Mode B: New Discovery	Sort by volume_per_hour
"What markets just opened?"	Mode B: New Discovery	Same as Mode B
"What crypto markets can I trade?"	Mode C: Category Browse (CRY)	Single category view
"What political markets are there?"	Mode C: Category Browse (POL)	Single category view
"What markets are available?" / "What can I trade?"	Mode C: Multi-category overview	No category → show all
"AI-related markets"	Mode C: Category Browse (TEC)	Keyword mapping
"Any sports markets to bet on?"	Mode C: Category Browse (SPT)	Single category view
"Fastest growing markets recently"	Mode B: New Discovery	Sort by volume_per_hour
"Which market has the most traders?"	Mode A: Trending	Sort by unique_traders_24h instead

Notes

Trade data window is ~2–7 days: All trade-based metrics (24h volume, trader count, new market detection) are limited by this. Queries beyond the data window return incomplete results.
"New market" is an approximation: Uses first trade time to approximate market launch. Market may have been created earlier with no trading activity.
Prices are real-time snapshots: Reflect the price at query time, not a period average.
Price field is probability, not dollars: 0.34 means the market believes 34% probability. A trader buys YES at $0.34, receives $1.00 if the event occurs.
Gamma API rate limit: Min 400ms between requests, max 20 condition_ids per batch.
condition_id format differs across systems: No 0x prefix in MCP SQL tables, needs 0x prefix for Gamma API queries.
Smart money attention is optional enrichment: Not required every time. Only check on new market discovery (top 5) and single category browse (top 1). Query failure doesn't affect main output.
Active market count = "markets with 24h trades": Not equal to total Polymarket market count. Many markets may have no recent trading.
User-facing output must NEVER expose internal implementation details: specific databases, provider APIs, MCP, analysis tables, and condition identifiers exist only in this SKILL.md internal workflow and must never appear in user-facing replies.