results-season — скилл для Claude Code: реакция акций на отчёты
Скилл results-season для Claude Code анализирует реакцию акций NSE и BSE на квартальные отчёты. Он оценивает гэп, объём и движение цены после публикации.
MIT ★ 69 обновлён 2026-11-13
/plugin marketplace add Shubhamkumartrichy/claude-trading-skills-india
/plugin install india-trading-skills@india-trading-skillsЧто делает results-season
Скилл разбирает движение акций NSE и BSE после квартальных отчётов: определяет торговую сессию реакции по времени публикации, считает гэп, доходность, объём относительно предыдущих 50 сессий и последующее движение. Затем присваивает оценку от 0 до 100 с буквенной категорией и весами факторов.
Команды анализируют одну акцию, сканируют список событий, проверяют гэп и отслеживают возможный вход после сильного роста. Скрипт results.py работает офлайн с CSV-файлами. Для расчёта 200-дневной средней нужно не менее 220 дневных баров; неоднозначную числовую дату скрипт отклоняет, а не угадывает.
Какие задачи решает
- оценить реакцию акции на квартальный отчёт
- сравнить реакцию на отчёты по списку событий
- проверить, заполнился ли ценовой гэп
- найти позиции с отчётами внутри срока удержания
Примеры запросов агенту
- Оцени реакцию TCS на отчёт за дату 2026-10-09 по файлу
tcs.csv. - Просканируй события из
q2.csvи ранжируй акции по реакции. - Проверь, у каких моих позиций отчёты попадают в период с 2026-11-02 по 2026-11-13.
Как установить results-season
1Через маркетплейс плагинов
/plugin marketplace add Shubhamkumartrichy/claude-trading-skills-india/plugin install india-trading-skills@india-trading-skillsКоманда ставит плагин india-trading-skills целиком, со всеми его скиллами
2Через Skills CLI (Vercel)
npx skills add Shubhamkumartrichy/claude-trading-skills-india --skill results-season -a claude-code3Вручную: положить файлы
~/.claude/skills/results-season/SKILL.mdПапка скилла целиком; для одного проекта — .claude/skills/results-season/
1Через Skills CLI (Vercel)
npx skills add Shubhamkumartrichy/claude-trading-skills-india --skill results-season -a codex2Вручную: положить файлы
.agents/skills/results-season/SKILL.mdДля всех проектов — ~/.agents/skills/
1Через Skills CLI (Vercel)
npx skills add Shubhamkumartrichy/claude-trading-skills-india --skill results-season -a cursor2Вручную: положить файлы
.cursor/skills/results-season/SKILL.mdCursor читает и .agents/skills/, и .claude/skills/
1GitHub CLI
gh skill install Shubhamkumartrichy/claude-trading-skills-india skills/results-season2Вручную: положить файлы
.github/skills/results-season/SKILL.md1Через Skills CLI (Vercel)
npx skills add Shubhamkumartrichy/claude-trading-skills-india --skill results-season -a gemini-cli2Командой платформы
gemini skills install https://github.com/Shubhamkumartrichy/claude-trading-skills-india --path skills/results-seasonВручную: положить файлы
.devin/skills/results-season/SKILL.mdСтарый путь .windsurf/skills/ тоже читается
Claude.ai
Упакуйте папку в zip и загрузите в Settings → Capabilities → Skills
По шагам, с командами для каждой платформы: Как установить скилл в Claude Code: пять способов
Вопросы о results-season
Как установить results-season в Claude Code?
Добавь маркетплейс командой /plugin marketplace add Shubhamkumartrichy/claude-trading-skills-india, затем установи скилл: /plugin install india-trading-skills@india-trading-skills.
Что делает results-season?
Он анализирует реакцию акций NSE и BSE на квартальные отчёты, оценивает движение цены и объёма, а также отслеживает гэп и последующее движение.
Что нужно для работы results-season?
Нужен Python 3 и CSV-файлы с данными по акциям. Для расчёта 200-дневной средней требуется не менее 220 дневных баров.
Исходный текст
SKILL.md · MIT
--- name: results-season description: | Reads the price reaction to quarterly results on NSE/BSE stocks. Picks the right reaction session from the announcement time (same day for pre-market/market-hours, next NSE session for after-hours, weekend or holiday releases), measures gap, close, close location, volume vs 50-day, ATR-normalised move, 1-20 session follow-through and gap fill, grades it 0-100 (A-D) with visible weights, watches weekly bars for a post-results drift entry or a failed reaction, ranks many events, and lists positions whose results fall inside a holding window. results.py works offline from CSVs. Use when the user asks: "how did X react to results", "was that a good results reaction", "should I buy after the results gap", "post-earnings drift setup in X", "rank this quarter's results reactions", "did the gap fill", "which of my holdings report before I exit", "results calendar check", or plans a swing trade during results season. --- # Results Season (NSE/BSE) `fundamental-screening` reads the numbers. This skill reads the market's verdict on them. The reaction (how far, on how much volume, where it closed, and whether it held) is more objective than a headline beat or miss, especially in India, where consensus estimates are not freely available. ## Commands ```bash # One stock: reaction metrics, then the score (adds factor table and a plain-English read) python3 scripts/results.py reaction --csv tcs.csv --date 2026-10-09 --timing post-market python3 scripts/results.py score --symbol HDFCBANK --date 18-Oct-2026 --timing 14:20 # Weekly drift watch after a strong gap-up: pullback that holds the gap-day low, then a breakout python3 scripts/results.py drift --csv dixon.csv --date 2026-07-24 --timing post-market # Many events: symbol,date,timing CSV (+ optional file column); CSVs named SYMBOL.csv python3 scripts/results.py scan --events q2.csv --csv-dir data/ --top 10 # Positions with results inside a holding window (feeds trading-discipline's event check) python3 scripts/results.py calendar-check --events q2.csv --from 2026-11-02 --to 2026-11-13 --symbols TCS,INFY ``` Inputs follow `technical-analysis` (`--csv`, `--candles`, `--symbol`). Dates are ISO (2026-10-31) or NSE style (31-Oct-2026); a numeric date that could be DD/MM or MM/DD is refused, never guessed. `--timing` is pre-market, market-hours, post-market, or the filing time (16:05, read as IST). Every command takes `--json`. Bad input exits 2 with one `error:` line. Use 220+ daily bars for the 200-DMA. ## Which session is the reaction? | Announced | Reaction session | |---|---| | Before 09:15 on a trading day | Same day: the open is the verdict | | 09:15–15:30 | Same day. The open came before the news, so read the close, not the gap | | 15:00–15:30 | Same day, flagged: NSE's close is the volume-weighted price of the last 30 minutes, so part of the move lands next session | | After 15:30, or a weekend/holiday | Next NSE session (holiday list from `../fno-options-analysis/scripts/options.py`) | Some companies hold results board meetings on Saturdays, and those react on Monday. Muhurat or special sessions in the data are skipped and flagged. Use `--reaction-date` to override. ## Reading `reaction` and `score` `reaction` reports gap %, reaction-day return, close location (0 = low, 1 = high), volume ÷ the prior 50 sessions, ATR(14) as of the day before, follow-through at +1/3/5/10/20 sessions (vs the reaction close and vs the pre-results close, and whether the reaction-day low held), and whether the gap filled. | Factor | Weight | Full marks | Why | |---|---|---|---| | Reaction size | 35 | close ≥ +3 ATR vs pre-results close | ATR scales the move to the stock's usual noise; the close is what the market kept | | Volume expansion | 20 | 4× the 50-day average (1× = 0) | Funds re-rating a stock can't hide their size | | Close strength | 15 | closed at the day's high | A faded gap means holders used the news to sell | | Trend position | 20 | above 200-DMA (7), above 50-DMA (6), at the pre-results 52-week high (7, zero at −10%) | Gaps into new highs meet no trapped sellers | | Fresh surprise | 10 | ≤ +10% in the 60 sessions before results (zero at +40%) | A big run-up often means the news was already priced in | Volume, trend and fresh surprise are context, so they are multiplied by a **gate**: 0 for a flat or negative close and 1 from +1 ATR up. A stock in a perfect uptrend that does nothing on results earns little. Missing data (no volume, under 200 bars) drops that factor and rescales the total to 100. Grades: **A ≥80, B ≥60, C ≥40, D below**. The score is long-biased. A strong negative reaction is a D with a bearish read. ## Reading `drift` Runs only after a strong gap-up: gap or close ≥ `--min-gap-pct` (4%) with a green close. Weeks start after the reaction week. - **Failed:** any daily close below the gap-day low. The reaction was rejected, so drop it. - **Pullback-set:** the first red weekly candle whose low stays above the gap-day low. The trigger is that week's high, and the stop is its low minus `--stop-buffer-pct` (0.5%). Further red weeks extend the pullback and lower the trigger. A week that dips under the gap-day low intraday but closes above it is noted as a shakeout. - **Triggered / stopped:** the later breakout above the trigger (a gap above it fills at the open), then the stop check. Reports R so far. **Expired:** no entry within `--max-weeks` (12). Feed the trigger and stop to `../breakout-planner/scripts/breakout.py plan --pivot .. --stop ..` for ticks, size and GTT orders. ## India context - **Deadlines (SEBI LODR Reg 33(3)):** Q1–Q3 results within 45 days of quarter end (14 Aug, 14 Nov, 14 Feb). Q4 comes with the audited annual results within 60 days of the FY end (30 May). `calendar-check` prints the season deadline and flags positions with no known date. - **Notice (Reg 29):** the board meeting to approve results must be intimated at least 2 working days ahead, excluding the intimation and meeting days. This has applied since the 17-May-2024 amendment; it was 5 days before that. Exchange board-meeting lists therefore cover only the next few days. Re-check them during long holds. - **Timing (Sch III Part A):** the outcome must be filed within 30 minutes of the meeting ending, so results can land mid-session as well as after 15:30. Use the filing timestamp. - **F&O stocks:** implied volatility builds into results and collapses after (IV crush). Option buyers can be right on direction and still lose. Compare the IV-implied move from `fno-options-analysis` with the realised ATR move here. - **Non-F&O stocks** have 5/10/20% price bands. A results-day upper circuit means buyers may not have been filled. A lower circuit traps holders, and stops do not fill. The script flags closes pinned at a band. - **No free consensus.** Judge the numbers by YoY growth first (seasonality), then QoQ, plus margins and management commentary (`fundamental-screening`). The reaction shows how they compared with what the market expected. ## How it connects - `india-market-analyst`: avoid fresh swing entries just before results. Run `calendar-check` before any multi-day plan. - `trading-discipline`: an INSIDE row from `calendar-check` means the gate gets `--event yes`. - `momentum-scanner`: a grade-A gap out of a quiet base is an episodic pivot candidate. - `breakout-planner`: a pivot within ~5 sessions of results is a coin toss. After the event, the drift trigger is a cleaner pivot. - `trailing-stops`: manage a triggered drift trade (breakeven at 1R). ## Limits These are daily bars only: no intraday order of events, no delivery %, no option IV. Reactions to other news on the same day (block deals, index inclusion, market-wide gaps) look like results reactions, so check the NIFTY move that day. Thresholds are conventions, not fitted parameters. _Educational analysis, not investment advice._
© Shubham Kumar · MIT · Shubhamkumartrichy/claude-trading-skills-india