korea-market

Name: korea-market
Author: kansoku-trade

226stars19forks

How to install korea-market?

kansoku-trade/kansoku/korea-market

$npx -y skills add kansoku-trade/kansoku --skill korea-market

Installs into the current project.

Summary

Korean market quotes (KOSPI / KOSDAQ / SK Hynix / Samsung) with capitulation-reversal detection. Use whenever reading the US memory/storage complex (MU / DRAM / SNDK / WDC / STX / SMH) — Korea is the SOURCE market and leads the US tape; Longbridge does not cover KRX. Triggers: 韩国

Files · 1

View on GitHub

SKILL.md

1	# Korea Market
2
3	Korean quotes + a capitulation-reversal test, from the Yahoo Finance chart API. Stdlib only, no third-party deps, goes through `_shared/client.py` (cache + throttle + output contract).
4
5	## Why this skill exists
6
7	Longbridge does not cover KRX. It returns `[]` for `000660.KS` / `005930.KS`.
8
9	The tempting workaround — read Korea through the US-listed proxies EWY (Korea ETF) and KORU (3× Korea) — lies to you:
10
11	- They are FX-contaminated (a won move shows up as a price move).
12	- They are frozen while Seoul is still trading — they only reprice during US hours.
13
14	The 2026-07-14 proof. Reading the proxies, EWY looked ~flat (−0.13%) — nothing to see. The real Seoul tape that same session:
15
16	\| \| Seoul (real) \| Proxy said \|
17	\| -------- \| -------------------------------------------------------------------------------------- \| ---------- \|
18	\| SK Hynix \| low −9.1% intraday → closed +2.9% on 1.55× volume (heaviest of the entire selloff) \| EWY "flat" \|
19
20	That was a textbook capitulation bottom, and the proxy hid it completely. Anyone watching EWY would have missed the exact session they were waiting for.
21
22	Rule: when the question is about Korea, read Korea.
23
24	## Usage
25
26	```bash
27	python3 .claude/skills/korea-market/scripts/quote.py --smoke # connectivity self-test
28	python3 .claude/skills/korea-market/scripts/quote.py # default: KOSPI, KOSDAQ, SK Hynix, Samsung
29	python3 .claude/skills/korea-market/scripts/quote.py --fresh # bypass 5-min cache (use intraday)
30	python3 .claude/skills/korea-market/scripts/quote.py 000660.KS # single name
31	python3 .claude/skills/korea-market/scripts/quote.py --range 6mo # longer history
32	```
33
34	Yahoo symbols: KRX stocks are `NNNNNN.KS` (SK Hynix `000660.KS`, Samsung `005930.KS`); indices are `^KS11` (KOSPI), `^KQ11` (KOSDAQ).
35
36	## The exhaustion evidence (the point of this skill)
37
38	A capitulation bottom BY DEFINITION prints a new low. Testing for "no new low for two sessions" — the obvious naive rule — skips the exact session you are waiting for. That mistake was made on 2026-07-13 and corrected the next day by the data.
39
40	Look for exhaustion instead: heaviest selling _into_ a new low, then buyers taking the other side.
41
42	\| Condition \| Field \| Reference \|
43	\| ---------------------------------------------------------------- \| --------------- \| ------------------------------------ \|
44	\| New low — sellers pushed below the prior floor \| `made_new_low` \| `low < prior low` \|
45	\| Heavy volume — selling was maximal, not a drift \| `heavy_volume` \| `rel_volume ≥ 1.3×` (20-session avg) \|
46	\| Green close — buyers won the session \| `green_close` \| `close > prev_close` \|
47	\| Closed strong — decisively, not on a last-minute bell bounce \| `closed_strong` \| `(close−low)/(high−low) ≥ 0.6` \|
48
49	A washout on light volume is not capitulation — it is a drift lower with nobody home, which has no natural floor. Volume is what separates _"sellers are done"_ from _"buyers left."_
50
51	### These are references, not a rule. You judge.
52
53	The script deliberately does not emit a verdict. It reports each measurement next to the reference it is being read against, plus a `conditions_cleared` tally — and stops there.
54
55	Why: collapsing this into a boolean throws away the distinction that actually matters. On 2026-07-14, Samsung cleared 3 of 4 with volume at 1.11× against a 1.3× reference — a near-miss. A hard threshold reads that as an identical outcome to KOSDAQ's 1 of 4 (still red, still no bid). Those are not the same animal, and a script that says so is lying to you.
56
57	Read the numbers. Weigh them against the tape. Then decide.
58
59	- `--min-rel-volume` / `--min-close-position` tune the references if a name's normal volume profile warrants it.
60	- Indices (`^KS11` / `^KQ11`) are context, not evidence. Index volume is a poor exhaustion gauge — it dilutes the memory names across ~900 constituents. Weight the individual names much more heavily.
61	- A single name reversing is not a sector washout. Look for agreement across SK Hynix and Samsung, and confirm on the next session before treating anything as a floor.
62
63	## Interpretation rules
64
65	- Korea leads, the US follows. Read the Seoul close _before_ forming

People who liked this also viewed

6 picks

Type

korea-market

226stars19forks

View on GitHub

How to install korea-market?

kansoku-trade/kansoku/korea-market

$npx -y skills add kansoku-trade/kansoku --skill korea-market

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Korea Market
2
3	Korean quotes + a capitulation-reversal test, from the Yahoo Finance chart API. Stdlib only, no third-party deps, goes through `_shared/client.py` (cache + throttle + output contract).
4
5	## Why this skill exists
6
7	Longbridge does not cover KRX. It returns `[]` for `000660.KS` / `005930.KS`.
8
9	The tempting workaround — read Korea through the US-listed proxies EWY (Korea ETF) and KORU (3× Korea) — lies to you:
10
11	- They are FX-contaminated (a won move shows up as a price move).
12	- They are frozen while Seoul is still trading — they only reprice during US hours.
13
14	The 2026-07-14 proof. Reading the proxies, EWY looked ~flat (−0.13%) — nothing to see. The real Seoul tape that same session:
15
16	\| \| Seoul (real) \| Proxy said \|
17	\| -------- \| -------------------------------------------------------------------------------------- \| ---------- \|
18	\| SK Hynix \| low −9.1% intraday → closed +2.9% on 1.55× volume (heaviest of the entire selloff) \| EWY "flat" \|
19
20	That was a textbook capitulation bottom, and the proxy hid it completely. Anyone watching EWY would have missed the exact session they were waiting for.
21
22	Rule: when the question is about Korea, read Korea.
23
24	## Usage
25
26	```bash
27	python3 .claude/skills/korea-market/scripts/quote.py --smoke # connectivity self-test
28	python3 .claude/skills/korea-market/scripts/quote.py # default: KOSPI, KOSDAQ, SK Hynix, Samsung
29	python3 .claude/skills/korea-market/scripts/quote.py --fresh # bypass 5-min cache (use intraday)
30	python3 .claude/skills/korea-market/scripts/quote.py 000660.KS # single name
31	python3 .claude/skills/korea-market/scripts/quote.py --range 6mo # longer history
32	```
33
34	Yahoo symbols: KRX stocks are `NNNNNN.KS` (SK Hynix `000660.KS`, Samsung `005930.KS`); indices are `^KS11` (KOSPI), `^KQ11` (KOSDAQ).
35
36	## The exhaustion evidence (the point of this skill)
37
38	A capitulation bottom BY DEFINITION prints a new low. Testing for "no new low for two sessions" — the obvious naive rule — skips the exact session you are waiting for. That mistake was made on 2026-07-13 and corrected the next day by the data.
39
40	Look for exhaustion instead: heaviest selling _into_ a new low, then buyers taking the other side.
41
42	\| Condition \| Field \| Reference \|
43	\| ---------------------------------------------------------------- \| --------------- \| ------------------------------------ \|
44	\| New low — sellers pushed below the prior floor \| `made_new_low` \| `low < prior low` \|
45	\| Heavy volume — selling was maximal, not a drift \| `heavy_volume` \| `rel_volume ≥ 1.3×` (20-session avg) \|
46	\| Green close — buyers won the session \| `green_close` \| `close > prev_close` \|
47	\| Closed strong — decisively, not on a last-minute bell bounce \| `closed_strong` \| `(close−low)/(high−low) ≥ 0.6` \|
48
49	A washout on light volume is not capitulation — it is a drift lower with nobody home, which has no natural floor. Volume is what separates _"sellers are done"_ from _"buyers left."_
50
51	### These are references, not a rule. You judge.
52
53	The script deliberately does not emit a verdict. It reports each measurement next to the reference it is being read against, plus a `conditions_cleared` tally — and stops there.
54
55	Why: collapsing this into a boolean throws away the distinction that actually matters. On 2026-07-14, Samsung cleared 3 of 4 with volume at 1.11× against a 1.3× reference — a near-miss. A hard threshold reads that as an identical outcome to KOSDAQ's 1 of 4 (still red, still no bid). Those are not the same animal, and a script that says so is lying to you.
56
57	Read the numbers. Weigh them against the tape. Then decide.
58
59	- `--min-rel-volume` / `--min-close-position` tune the references if a name's normal volume profile warrants it.
60	- Indices (`^KS11` / `^KQ11`) are context, not evidence. Index volume is a poor exhaustion gauge — it dilutes the memory names across ~900 constituents. Weight the individual names much more heavily.
61	- A single name reversing is not a sector washout. Look for agreement across SK Hynix and Samsung, and confirm on the next session before treating anything as a floor.
62
63	## Interpretation rules
64
65	- Korea leads, the US follows. Read the Seoul close _before_ forming

People who liked this also viewed

6 picks

Type