agentleFS
ShubhamkumartrichyShubhamkumartrichy/claude-trading-skills-india/skills/exhaustion-reversal/SKILL.md

exhaustion-reversal

Skill64 starsChanged 17 days agoMITView on GitHub

When a move on NSE/BSE has run out of buyers or sellers. exhaustion.py scores a vertical run-up 0-100 (stretch above the 20/50/200 DMA, acceleration, run length, range/volume blow-off, late gaps, upper-circuit streaks, price-band cuts), labels it extended, climax or reversal confirmed, and plans the exit: scale out and trail with a GTT if you hold, or short only via stock futures or puts on F&O names after a trigger, sized in lots. Also finds selling-exhaustion hammers for long entries, judges whether a stock failed to react to news, and scans a universe with F&O tags.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    ASM/GSM and band cuts: After sharp run-ups exchanges move stocks into surveillance, cut the band 20 → 10 → 5 → 2% and demand 100% margin.

    line 115
  2. Very specific:

    Run the trading-discipline gate (../trading-discipline/scripts/discipline.py) first.

    line 130

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Exhaustion & Reversal (NSE/BSE)

momentum-scanner finds stocks as they start to move. This skill takes the same stocks weeks later, when the move is old and steep, and asks whether it is ending. For most Indian retail traders the answer matters as a holder: how much to sell into strength and where to trail the rest. Shorting is the exception, and only possible in F&O stocks.

Commands

python3 scripts/exhaustion.py climax --csv data/KAYNES.csv              # one stock: factors + stage
python3 scripts/exhaustion.py climax --csv-dir data/ --min-score 40     # ranked table
python3 scripts/exhaustion.py plan --csv data/KAYNES.csv --qty 120 --avg-price 4100
python3 scripts/exhaustion.py plan --csv-dir data/ --symbol KAYNES --side short --fno-list fo_mktlots.csv \
    --ban-list fo_secban.csv --margin-per-lot 210000 --capital 1500000 --risk-pct 1 --put-premium 120 --put-strike 5000
python3 scripts/exhaustion.py washout --csv-dir data/ --days 3 --capital 500000 --show-rejects
python3 scripts/exhaustion.py reaction --csv data/INFY.csv --date 2026-10-16 --news bullish --after-close \
    --crowding long --benchmark-csv nifty.csv
python3 scripts/exhaustion.py scan --csv-dir data/ --fno-list fo_mktlots.csv --bands bands.csv

Universe flags match momentum.py: --csv-dir + --pattern, --csv (repeatable), --symbols (yfinance), --as-of, --min-turnover-cr (5), --bands (SYMBOL,band, 0 for F&O), --tick, --json. --fno-list takes one symbol per line, SYMBOL,LOT or NSE fo_mktlots.csv (lots are read from it). Blank or zero candles are skipped; bad input exits 2 with one error: line.

climax: the parabolic score

The high of the move is the highest high in the last 15 sessions (--lookback). Points:

FactorMaxFull marks at
Stretch25high 6+ ATR over the 20 DMA (10), 50%+ over the 50 DMA (10), 120%+ over the 200 DMA (5)
Acceleration15last 5 sessions running 3x the 20-day daily pace, with 10%+ in those 5
Run25+50% in 10 / +80% in 20 / +150% in 60 sessions (15), 8 up days (6) and 6 up weeks (4) in a row
Blow-off30widest bar 3x ATR and volume 4x the 50-day average, each the biggest in 60 sessions, gap-ups in the last 5 sessions, and a peak day that closes in its lower part (reversal bar). Or 6 points per consecutive upper-circuit close (5 locked closes = 30): whichever is higher
Band cut5the exchange narrowed the price band during the run (or the stock sits in a 2%/5% band)

Why these: stretch and run size say how far price has left value; acceleration and up-day streaks catch the vertical last leg; the blow-off bar is where late buyers and short-covering arrive at once. Banded small caps can't print a blow-off bar (they lock), so the circuit streak stands in.

Stage: under 40 not extended, 40-64 extended, 65+ climax, and reversal confirmed once any of these happens after the high: a close below the previous session's low, a close below the 10 EMA, or a lower high (a bounce that fails under the peak, then a close under the pullback low). The date of the first is printed. Until then the output lists tomorrow's levels to watch.

plan

Holder (default --side hold). Sell a slice into strength: 25% when extended, 40% at a climax, 50% once confirmed (--sell-pct). Trail the rest one tick under the prior day's low (climax or a circuit streak) or on a close below the 10 EMA (extended), as a GTT single SELL with a limit 1-2% under the trigger. GTTs fire on LTP and send a limit, so a gap below the limit stays unfilled; plain SL orders expire at 15:30. On a locked upper circuit, selling into the buy queue fills at once, which is the best liquidity a pumped stock will offer. In a 2%/5% band a lower-circuit lock blocks every exit. See trailing-stops and stop-loss-strategies for the general mechanics.

Short (--side short or both). Refused for a non-F&O stock: cash shorts are intraday (MIS) only, an undelivered short goes to the exchange auction with a penalty, and there is no future or put to hold overnight. Also refused in the F&O ban (OI above 95% of MWPL: reduce-only) and below score 40 (--force). Otherwise: entry only after a trigger, a trade below the climax-day low or a failed bounce, never into strength. Stop is 0.5% above the climax high (--stop-buffer-pct). Targets are the 10 EMA, 20 EMA and 50 DMA with their R multiple. Size with --capital --risk-pct in whole lots (--lot, or from --fno-list). Margin is --margin-per-lot from your broker, else 25% of contract value, capped at 50% of capital. --put-premium/--put-strike prices a long put as the defined-risk route (see fno-options-analysis, options-strategies). Every short plan prints the warnings: no ceiling on loss, squeeze into expiry and MWPL, gap-through stops, physical settlement.

washout: selling exhaustion for a long entry

A bar in the last 3 sessions (--days) that is the low of the last five and: a pullback of 8-40% from the 40-session high over 3+ sessions; a prior uptrend (+25% into the high, a rising 200 DMA, or a name on your --quality list); a wick of 50%+ of the bar and a close in the top 40%; range ≥ 0.75 ATR; volume ≥ the 50-day average. Score: context 20, depth 15, undercut of an earlier pivot low then a close back above it 25, wick 15, close 10, volume 15, +5 when a lower-circuit streak just ended. Entry one tick over the bar's high, stop one tick under its low (NSE ticks), targets the 20 EMA and the prior high, status pending / triggered / failed.

reaction: did the news work?

Day 1 is the news date, or the next session with --after-close. Moves are measured in ATRs with the news (net of --benchmark-csv if given) over --sessions (3). Failed if day 1 closes against the news in the worst 35% of its range on 1.5x+ volume, or if after 2-3 sessions price is 0.5+ ATR the wrong way and through the news-day low (high, for bad news). Confirmed if 0.5+ ATR the right way and holding the news-day midpoint. Otherwise inconclusive. The rule that fired is printed. --crowding long|short|none adds a gate: crowded side + failed reaction + broken level = a contrarian setup. Judge crowding from fii-dii-flows (participant OI), the stock's OI vs MWPL, and MTF/delivery build-up.

scan

Climax scores ≥ 40 and washout bars, ranked, illiquid names filtered (listed with the reason), each tagged with its short route: futures/puts, F&O ban, or none overnight.

Reading it in India

  • Operator small caps and SME stocks rise in circuit streaks on thin delivery, no news. The queue that locked them up becomes the queue that locks them down. Sell into the circuit; never short them (you can't overnight, and intraday is usually blocked).
  • ASM/GSM and band cuts. After sharp run-ups exchanges move stocks into surveillance, cut the band 20 → 10 → 5 → 2% and demand 100% margin. Leverage leaves and the run often ends. Check the NSE/BSE surveillance lists and series (EQ vs BE).
  • Results-day spikes gap 10-20% in F&O names. A spike that closes near its low on record volume is a reversal bar; reaction tells you whether good results were sold.
  • Why shorting strength is dangerous: parabolic stocks go further than anyone expects, F&O names have no circuit, MWPL squeezes force covering, and futures are marked to market daily. The defined-risk way is a put or bear put spread, after confirmation.

Workflow

  1. climax on what you hold (or scan). Under 40: let trailing-stops handle it.
  2. Extended or climax: plan → sell the slice, set the GTT. Confirmed: sell the bigger slice.
  3. Short only F&O names, after the trigger, sized by position-sizing rules, 1% risk or less.
  4. washout entries: size with position-sizing (../position-sizing/scripts/risk.py).
  5. Run the trading-discipline gate (../trading-discipline/scripts/discipline.py) first.

Limits: daily bars only (no intraday triggers or live circuit status); bands are inferred unless --bands is given; F&O status, lots, ban and margin are only what you pass; the score is a description of the tape, not a probability that the top is in.

Educational analysis, not investment advice.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.