| name | watchlist-scan |
| description | scan configured polymarket watchlists for material changes. use when the user wants recurring monitoring, threshold-based alerts, catalyst watch, or a scheduled automation that reports only when something important changed. use opportunity-classifier next when the goal is to rank, segment, or filter the changed names. |
watchlist scan
Use this skill to answer one question:
what changed in the watchlist that actually deserves attention right now?
This skill is the monitoring layer:
- downstream of
configs/watchlists.yaml, bookmark sync, and persisted state
- upstream of
opportunity-classifier, market-memo, and strategy-draft
It should report meaningful deltas, not restate the whole watchlist.
default mode
- Default to delta mode.
- In delta mode, optimize for threshold breaches, new alerts, catalyst-adjacent movement, and workflow routing.
- Do not turn the scan into a generic market roundup.
what this skill is for
Use this skill when the user is effectively asking:
- what moved in my watchlist?
- which names had meaningful alerts?
- did any bookmarked or watched markets get more interesting?
- what should I look at first after a monitoring run?
Use this skill for:
- recurring scans
- threshold-based alert summaries
- change-focused reporting
- deciding which changed names deserve memo, research, or strategy work next
Do not use this skill to:
- rank an entire discovery universe
- write full market memos on every changed name
- place or cancel orders
- fill the report with unchanged markets
minimum viable input checklist
Do not finalize a scan until you have, or explicitly mark as missing:
- a watchlist source
- the latest scan state or alert state
- at least one threshold or condition for materiality
- the changed markets that crossed those thresholds, if any
Prefer to build on persisted state when available.
outside-in frame
Before reporting, inspect the scan from these lenses:
- change lens: what actually changed, not just what exists?
- materiality lens: is the move large enough, timely enough, or weird enough to matter?
- workflow lens: should a changed name go to classifier, memo, research, or strategy next?
- noise lens: what can be safely omitted because it is unchanged or trivial?
If the noise lens says nothing matters, say so briefly and stop.
dependency and handoff rules
Use the minimum monitoring set needed:
- load
configs/watchlists.yaml
- prefer
get_live_alerts, get_state_summary, or get_market_state when local SQLite state is available
- use fresh snapshot reads only where the alert or state data is insufficient
If the main missing piece is:
- which changed names deserve deeper prioritization ->
opportunity-classifier
- compact current-state read on one changed name ->
market-memo
- fair-value or catalyst work on one changed name ->
deep-market-research
- execution framing after a changed setup becomes actionable ->
strategy-draft
main workflow
1. Normalize the assignment
Write a 1-2 line intent statement for yourself covering:
- which watchlist source is in scope
- whether the user wants a one-off scan or an automation-style summary
- what counts as material in this run
2. Load the watchlist and prior state
Start with:
Then prefer:
get_live_alerts
get_state_summary
get_market_state
Use current-thread-only scans only when persisted state is unavailable.
3. Filter to material changes
Only surface names that:
- crossed configured thresholds
- generated fresh alerts
- showed meaningful related-market drift
- developed a new comment or catalyst signal that materially changes attention level
Ignore unchanged names and trivial noise.
4. Summarize the change, not the entire market
For each surfaced name, state:
- what changed
- why it matters
- what next step is justified
Keep each name compact unless the user asks to drill down.
5. End with workflow routing
For each changed name, route to the cheapest useful next step:
opportunity-classifier
market-memo
deep-market-research
strategy-draft
- no action
hard rules
- Keep the scan concise and change-focused.
- Do not place, preview, submit, or cancel orders in this skill.
- Do not promote every moved market as actionable.
- Do not turn the scan into a memo for unchanged or low-signal names.
- Do not omit "nothing material" when that is the honest result.
output contract
Return exactly these sections in this order.
watchlist scan
scan intent
State in 1-2 sentences:
- what watchlist source was scanned
- what materiality standard was used
material changes
If there are changes, use one flat bullet block per changed market:
- market name
- what changed
- why it matters
- recommended next step
nothing material
Use this section only when nothing crossed the reporting threshold.
State that briefly and stop.
next handoff
Choose one:
- remain in
watchlist-scan
- hand off changed names to
opportunity-classifier
- hand off one name to
market-memo
- hand off one name to
deep-market-research
- hand off one name to
strategy-draft
Explain why that is the cheapest useful next step.
completion criteria
The scan is complete only when:
- the watchlist source is explicit
- only material changes are surfaced
- unchanged noise is omitted
- each surfaced name has a why-it-matters note
- the next step is explicit
- no live trading action is taken
common failure modes
- dumping the whole watchlist instead of the deltas
- treating every price move as meaningful
- ignoring stored alert state and recomputing everything from scratch
- writing mini memos for each name without need
- failing to say "nothing material" when nothing actually changed
activation tests
Should trigger
- "scan my watchlist for anything important."
- "what changed in the watched markets?"
- "give me the alert-level summary from the watchlist."
- "tell me which watchlist names deserve a closer look after the latest move."
Should not trigger
- "classify these 20 markets."
- "write a memo on this one market."
- "research fair value on this event."
- "place an order."
- "review my portfolio risk."
resource map
opportunity-classifier — use when multiple changed names need ranking or filtering
market-memo — use when one changed name needs a compact live-state read
deep-market-research — use when the change justifies a full fair-value pass
strategy-draft — use when a watched market became actionable and now needs execution framing