Convert Kanchi-style dividend investing into a repeatable US-stock operating procedure. Use when users ask for かんち式配当投資, dividend screening, dividend growth quality checks, PERxPBR adaptation for US sectors, pullback limit-order planning, or one-page stock memo creation. Covers screening, deep dive, entry planning, and post-purchase monitoring cadence.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
La commande reste sur une seule ligne. Faites défiler horizontalement pour la vérifier avant de la copier.
Vous préférez une copie locale ? Téléchargez les fichiers actuellement disponibles dans SkillsMP.
Explorateur de fichiers
25 fichiers
Affichage de SKILL.md
SKILL.md
Instructions source · Aperçu en lecture seule
name
kanchi-dividend-sop
description
Convert Kanchi-style dividend investing into a repeatable US-stock operating procedure. Use when users ask for かんち式配当投資, dividend screening, dividend growth quality checks, PERxPBR adaptation for US sectors, pullback limit-order planning, or one-page stock memo creation. Covers screening, deep dive, entry planning, and post-purchase monitoring cadence.
Kanchi Dividend Sop
Overview
Implement Kanchi's 5-step method as a deterministic workflow for US dividend investing.
Prioritize safety and repeatability over aggressive yield chasing.
When to Use
Use this skill when the user needs:
Kanchi-style dividend stock selection adapted for US equities.
A repeatable screening and pullback-entry process instead of ad-hoc picks.
One-page underwriting memos with explicit invalidation conditions.
A handoff package for monitoring and tax/account-location workflows.
Prerequisites
API Key Setup
The entry signal script requires FMP API access:
export FMP_API_KEY=your_api_key_here
Input Sources
Prepare one of the following inputs before running the workflow:
Output from skills/value-dividend-screener/scripts/screen_dividend_stocks.py.
Output from skills/dividend-growth-pullback-screener/scripts/screen_dividend_growth_rsi.py.
User-provided ticker list (broker export or manual list).
Expected JSON Input Format
When using --input, provide JSON in one of these formats:
{"profile":"balanced"
,
"candidates"
:
[
{
"ticker"
:
"JNJ"
,
"bucket"
:
"core"
}
,
{
"ticker"
:
"O"
,
"bucket"
:
"satellite"
}
]
}
Or simplified:
{"tickers":["JNJ","PG","KO"]}
The optional value-dividend-screener and
dividend-growth-pullback-screener handoffs use stocks[].symbol.
Both build_sop_plan.py --input and build_entry_signals.py --input
accept that shape directly, as well as the native candidates[].ticker and
tickers[] shapes above.
For deterministic artifact generation, provide tickers to:
For Step 5 entry timing artifacts. --yield-floor is mandatory — it is
the Step-1 yield gate; without it every row fail-safes to STEP1-RECHECK
(a row can never reach a PASS tier without Step 1). Pass --profile /
--safety-bias for run_context, and --events-json for the Step 4b scan
(absent ⇒ every row is treated as SKIPPED and a TRIGGERED name is capped
to HOLD-REVIEW — never silently clean):
User-provided broker export or manual ticker list when APIs are unavailable.
Return a ticker list grouped by bucket before moving forward.
3) Apply Kanchi Step 1 (yield filter with trap flag)
Primary rule:
Step-1 yield = the regular forward yield = latest_declared_regular dividend × cadence-implied frequency / price (WS-1 dividend_basis.py).
Never use profile.lastDividend / TTM — it lags the latest declared raise
(defect D5) and silently bundles specials (D4).
Apply the profile floor (income-now 4.0% / balanced 3.0% / growth-first
1.5%) to the regular yield only.
Trap & freshness controls (machine-emitted by dividend_basis.py):
special_dividend_flag → exclude specials; report regular vs ttm yield.
variable_policy_flag → FAIL (CALM-style; not an income base).
cut_flag → FAIL; suspension_flag → FAIL.
freeze_flag → HOLD-REVIEW (income cash-cow exception decided in
Step 8 synthesis only if safety is clean & unblocked).
Data Freshness Gate: if the regular yield is within ±0.20pp of the
floor (floor_borderline) and the latest declared dividend is not
confirmed from an authoritative source, emit STEP1-RECHECK — never a
hard FAIL (this is the CFR D5 fix).
4) Apply Kanchi Step 2 (growth and safety) — sector-dispatched
Safety is sector-specific — a uniform GAAP/FCF triad mis-judges banks
(FCF meaningless) and regulated utilities (FCF structurally negative).
Use references/sector-step2-modules.md; the deterministic dispatch is
scripts/payout_safety.py.
Always compute the payout triad: GAAP-EPS payout, Adjusted-EPS
payout, FCF payout. The safety verdict uses Adjusted-EPS + FCF
(consumer), or the sector module (bank / utility / insurer).
adjusted_eps_source = UNAVAILABLE ⇒ cap HOLD-REVIEW (fail-safe;
never a silent PASS).
Step 4 is backward-looking; Step 4b catches pending/recent structural
events (the MKC-Unilever miss, D3). For each surviving candidate, run a
WebSearch + issuer-IR/SEC check using the source hierarchy: issuer IR
→ SEC filing (8-K/10-Q/10-K/proxy/S-4) → exchange/company deck →
reputable wire → finance portals (secondary only). Record findings into a
curated events JSON and pass it via build_entry_signals.py --events-json.
Only a major structural event caps the verdict to HOLD-REVIEW
(tx > 10% mcap, share issuance > 10–20%, leverage +0.5x EBITDA,
control/listing/HQ change, merger-of-equals / RMT / spin-off / large
asset sale, dividend/rating/leverage-policy change, sector-specific
materiality, or rolling-24m cumulative M&A > 15% mcap). Minor bolt-ons
are a CAUTION note only.
Pessimistic cap: FAILED-DEGRADED / SKIPPED / NO_EVENT_FOUND
on a Step-5 TRIGGERED name ⇒ HOLD-REVIEW + T1 BLOCKED. WebSearch
unavailable (web app / offline) is treated the same — never a silent
skip. CLEAN_CONFIRMED (primary source checked) is stronger than
NO_EVENT_FOUND (search only).
7) Apply Kanchi Step 5 (buy on weakness with rules)
Set entry triggers mechanically:
Yield trigger: current yield above 5y average yield + alpha (default +0.5pp).
Valuation trigger: target multiple reached (P/E, P/FFO, or P/FCF).
Execution pattern:
Split orders: 40% -> 30% -> 30%.
Pre-order blockers: if a candidate has any unresolved
pre_order_blockers[] (from WS-1/2/3 — variable/cut/suspension,
adjusted-EPS-unavailable, GAAP/Adj divergence, bank credit, utility
FFO/debt, event-scan failed/skipped, stale dividend, …) OR
t1_blocked is true, the first tranche is blocked or downsized to a
≤20% tracking tranche — not 40%.
Sector cluster risk: when ≥ SECTOR_CLUSTER_WARN_COUNT same-sector
names pass (e.g. many small banks share one macro beta), emit a
portfolio-level CLUSTER-RISK warning.
Require one-sentence sanity check before each unblocked add: "thesis
intact vs structural break".
8) Produce standardized outputs
Always produce:
Screening table with the actionable verdict tier: CLEAN-PASS,
PASS-CAUTION, CONDITIONAL-PASS, HOLD-REVIEW, STEP1-RECHECK,
FAIL (synthesized by verdict.py from Step 1 + Step 2 + Step 4b +
blockers). Include evidence per row.
One-page stock memo (use references/stock-note-template.md) with the
per-ticker provenance block (price/dividend/payout/event sources,
unresolved_blockers, evidence_refs[]).
Limit-order plan with split sizing, blocker gate, and invalidation.
Top-level run_context (profile, yield_floor_pct, safety_bias,
universe_source, excluded_asset_types) so a 3%-run result is never
silently reused inside a 4%-run.
Output
Return and/or generate:
SOP screening summary in markdown.
Underwriting memo set based on
references/stock-note-template.md.
Optional plan artifact file generated by
skills/kanchi-dividend-sop/scripts/build_sop_plan.py in reports/.
Optional Step 5 entry-signal artifacts generated by
skills/kanchi-dividend-sop/scripts/build_entry_signals.py in reports/.
Cadence
Use this minimum rhythm:
Weekly (15 min): check dividend and business-news changes only.
Monthly (30 min): rerun screening and refresh order levels.
Quarterly (60 min): deep safety review using latest filings/earnings.
Multi-Skill Handoff
Run this skill first, then hand off outputs:
To kanchi-dividend-review-monitor for daily/weekly/quarterly anomaly detection.
To kanchi-dividend-us-tax-accounting for account-location and tax classification planning.
Guardrails
Do not issue blind buy calls without Step 4, Step 4b and safety checks.
Do not treat high yield as value before validating coverage quality.
Use the regular forward yield for Step 1, never a special/TTM-inclusive
figure; near-floor + unconfirmed ⇒ STEP1-RECHECK, not FAIL.
A failed/skipped event scan on a TRIGGERED name ⇒ HOLD-REVIEW + T1
blocked. Never silently skip Step 4b.
