Watchlist Monitor
Read-only. reference/READ-ONLY-POLICY.md (hard rule) and
reference/RESEARCH-STANDARDS.md (data efficiency, news freshness,
completeness checklist, disclosure) apply. Steps name capabilities —
resolve each against reference/BROKER-CAPABILITIES.md.
Built to be run often, so it must stay cheap: one batched price call for
the whole list, and depth only on what's actually in range.
market-pulse covers what the market did; this covers what the user
said they were waiting for.
Steps
Read the watchlist. PORTFOLIO-PLAN.md — Watchlist / themes
(name, why interested, trigger), plus Exclusions, Risk limits
and Deployable capital (a hit with no cash behind it is a
different conversation), and the Decision log so a name the user
passed on isn't re-pitched as news. Empty watchlist → say so and offer
portfolio-plan-builder. Never substitute a generic idea list for
names the user actually chose.
Resolve each trigger into something testable: a price level, a
valuation level (P/E, yield, market cap), an event ("after Q2
results", "once the lock-in expires"), or a theme criterion ("any
index ETF under 0.5% TER"). Anything vaguer is a finding — say it
can't be checked and propose a concrete rewrite.
Price the list in one call. Symbol-search capability once for
unresolved identifiers, then a single batched LTP call for every
price-level name. The core output is distance to trigger in % —
compute it for all of them, sort by it, and stop there for anything
far away. Valuation triggers: fundamentals capability for those names
only. Theme triggers aren't screened here — hand them to
new-investment-screener, which is built for it.
Go deeper only on hit or near (~5%). Technical indicators (batched)
and 1–2 date-anchored WebSearch calls per name. The question worth
answering is why the level arrived: a de-rating on real news is not
the same opportunity as ordinary drift toward a number the user picked
months ago. Say which it looks like.
Check a hit against the rest of the plan before treating it as
actionable — pull equity holdings and test for existing exposure or
overlap (an "opportunity" the user already owns twice over isn't one),
single-stock and sector headroom under Risk limits, exclusions,
and cash under Deployable capital.
Present. One table sorted by distance: name, trigger, current
value, distance, status (hit / near / far), one-line note. Prose only
for hit and near names. A hit that carries a view needs the
completeness checklist — including that the trigger firing is not
itself a thesis — and the disclosure block; a sweep where nothing hit
carries no view and needs no disclosure. Data as-of near the top.
Formal version: reference/templates/watchlist-monitor.md.
Hand off. stock-research to actually underwrite a name that hit,
new-investment-screener for theme rows, earnings-watch for
results-dependent triggers, etf-tracking-quality when the candidate
is an index vehicle. Propose watchlist edits (a fired trigger to
retire, a stale row) for portfolio-plan-builder to write.
1---2name: watchlist-monitor3description: Sweep the watchlist in PORTFOLIO-PLAN.md — which triggers are hit, which are near, and what changed on the names now in range. Use when the user asks about their watchlist, whether anything hit their price, or "anything worth looking at today". Read-only.4---56# Watchlist Monitor78Read-only. `reference/READ-ONLY-POLICY.md` (hard rule) and9`reference/RESEARCH-STANDARDS.md` (data efficiency, news freshness,10completeness checklist, disclosure) apply. Steps name capabilities —11resolve each against `reference/BROKER-CAPABILITIES.md`.1213Built to be run often, so it must stay cheap: one batched price call for14the whole list, and depth only on what's actually in range.15`market-pulse` covers what the *market* did; this covers what the user16said they were waiting for.1718## Steps19201. **Read the watchlist.** `PORTFOLIO-PLAN.md` — **Watchlist / themes**21 (name, why interested, trigger), plus **Exclusions**, **Risk limits**22 and **Deployable capital** (a hit with no cash behind it is a23 different conversation), and the **Decision log** so a name the user24 passed on isn't re-pitched as news. Empty watchlist → say so and offer25 `portfolio-plan-builder`. Never substitute a generic idea list for26 names the user actually chose.27282. **Resolve each trigger into something testable**: a **price level**, a29 **valuation level** (P/E, yield, market cap), an **event** ("after Q230 results", "once the lock-in expires"), or a **theme criterion** ("any31 index ETF under 0.5% TER"). Anything vaguer is a finding — say it32 can't be checked and propose a concrete rewrite.33343. **Price the list in one call.** Symbol-search capability once for35 unresolved identifiers, then a single batched LTP call for every36 price-level name. The core output is **distance to trigger in %** —37 compute it for all of them, sort by it, and stop there for anything38 far away. Valuation triggers: fundamentals capability for those names39 only. Theme triggers aren't screened here — hand them to40 `new-investment-screener`, which is built for it.41424. **Go deeper only on hit or near (~5%).** Technical indicators (batched)43 and 1–2 date-anchored `WebSearch` calls per name. The question worth44 answering is *why* the level arrived: a de-rating on real news is not45 the same opportunity as ordinary drift toward a number the user picked46 months ago. Say which it looks like.47485. **Check a hit against the rest of the plan before treating it as49 actionable** — pull equity holdings and test for existing exposure or50 overlap (an "opportunity" the user already owns twice over isn't one),51 single-stock and sector headroom under **Risk limits**, exclusions,52 and cash under **Deployable capital**.53546. **Present.** One table sorted by distance: name, trigger, current55 value, distance, status (hit / near / far), one-line note. Prose only56 for hit and near names. A hit that carries a view needs the57 completeness checklist — including that the trigger firing is not58 itself a thesis — and the disclosure block; a sweep where nothing hit59 carries no view and needs no disclosure. Data as-of near the top.60 Formal version: `reference/templates/watchlist-monitor.md`.61627. **Hand off.** `stock-research` to actually underwrite a name that hit,63 `new-investment-screener` for theme rows, `earnings-watch` for64 results-dependent triggers, `etf-tracking-quality` when the candidate65 is an index vehicle. Propose watchlist edits (a fired trigger to66 retire, a stale row) for `portfolio-plan-builder` to write.
getlost01