agentleFS
ShubhamkumartrichyShubhamkumartrichy/claude-trading-skills-india/skills/stock-screener/SKILL.md

stock-screener

Skill64 starsChanged 17 days agoMITView on GitHub

Filter and rank a universe of Indian stocks from a fundamentals CSV (a Screener.in screen or watchlist export, or your own snake_case file). Five built-in screens with India thresholds: canslim-india, value-dividend (yield bar tied to the 10-year G-sec), dividend-growth-pullback (RSI from price CSVs), garp and quality-compounder. Also a safe --where filter language, near-miss lists, a per-company rule breakdown and the matching Screener.in query text. Red-flag gates: promoter pledge, ₹500 cr market-cap floor, negative net worth, missing data, and bank/NBFC handling. screen.py works offline.

12 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 0 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run a screen (or open a watchlist) on screener.in.

    line 42
  2. Very specific:

    Add an Industry column if you can: it is how banks and NBFCs are recognised.

    line 45
  3. Very specific:

    Read each passer's filings, governance and red flags with fundamental-screening.

    line 110
  4. Very specific:

    Before buying, check fit and concentration with portfolio-review.

    line 113

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “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

Stock Screener (India)

fundamental-screening explains how to read one company. This skill is the calculator: it runs a whole universe through explicit rules, shows each pass/fail with the value, and ranks what survives. A screen produces a reading list, not a buy list.

Commands

python3 scripts/screen.py presets                        # the five screens, rules, weights, reasons
python3 scripts/screen.py fields --csv export.csv        # how your headers mapped; field names
python3 scripts/screen.py run --csv export.csv --preset quality-compounder
python3 scripts/screen.py run --csv export.csv --where "roce>20" --where "pe<25" --where "pledge<=5" --sort roce --top 20
python3 scripts/screen.py run --csv export.csv --preset dividend-growth-pullback --prices-dir data/
python3 scripts/screen.py explain "Infosys" --csv export.csv --preset garp
python3 scripts/screen.py query --preset value-dividend --gsec 7.15

Common flags: --json (strict JSON, missing values are null), --gsec (current 10-year G-sec yield, default 7.0), --min-mcap (₹ crore, 500), --max-pledge (%, 20), --no-gates, --map "field=Header" (repeatable), --financials auto|none|NAME,NAME. A preset and --where can be combined. Bad input exits 2 with a one-line error:.

Getting the CSV out of Screener.in

  1. Run a screen (or open a watchlist) on screener.in. Use Edit Columns to add what the preset needs; screen.py presets and the "NOT CHECKED" notes tell you which.
  2. Click Export and save as CSV (if you get an .xlsx, open it and save the sheet as CSV).
  3. Add an Industry column if you can: it is how banks and NBFCs are recognised.

Headers are matched by keyword, so units, %, commas (1,23,456.70), ₹, blanks, - and NA are fine. Verified on public Screener.in screen tables (Nov 2026): S.No., Company, CMP Rs., P/E, Mar Cap Rs.Cr., Div Yld %, NP Qtr Rs.Cr., Qtr Profit Var %, Sales Qtr Rs.Cr., Qtr Sales Var %, ROCE %, ROE %, Debt / Eq, CMP / BV. Assumed (column names vary with what you add): EPS, 3/5-year growth, promoter/pledge/FII/DII holding and change, PEG, payout, dividend growth, free cash flow, 52-week high, returns. We have only seen the web tables, not an exported file, so if a header is not recognised, fields --csv lists it as unmapped. Fix it with --map pledge="Pledged %". Snake_case files work too: name,symbol,industry,price,market_cap_cr,pe, pb,dividend_yield,payout_ratio,roce,roe,debt_to_equity,profit_growth_3y,promoter_holding,pledge_pct.

Gates (every run unless --no-gates)

GateDefaultWhy
Market cap≥ ₹500 crBelow this: SME boards, thin floats, operator stocks, ASM/GSM lists
Promoter pledge≤ 20%Red-flag line in fundamental-screening; 10–20% is flagged pledge
Net worthnot negativeNegative book value, P/B or D/E means losses have eaten the equity
Missing dataany rule's value blankListed as INCOMPLETE, never silently passed (blank P/E usually = loss)

Banks, NBFCs, insurers (from the Industry column, otherwise guessed from the name) skip the debt, ROCE, free-cash-flow, interest-cover and margin rules in presets, and are tagged FIN. Judge them on ROA (> 1.2%), GNPA/NNPA, NIM and capital adequacy with fundamental-screening. Your own --where rules apply to them unchanged. Soft flags: small-cap (< ₹2,000 cr) and low-base? (quarterly profit up 200%+).

Presets (how each is adapted for India)

  • canslim-india: C = quarterly profit ≥ 25% YoY with sales ≥ 15% (no free consensus estimates in India, so reported growth stands in for a surprise). A = 3-year profit CAGR ≥ 20% and ROE ≥ 15%. N = within 15% of the 52-week high. L = 1-year return in the top 30% of your file, so export NIFTY 500 or wider. I = FII + DII holding rose last quarter (fii-dii-flows for detail). M is not in a fundamentals file: run ../market-regime/scripts/regime.py status (market-regime) first.
  • value-dividend: P/E ≤ 20 (NIFTY's long-run average is about 20–22), P/B ≤ 3, D/E ≤ 0.5, payout ≤ 75%, 3-year profit growth ≥ 5% (no yield traps), ROE ≥ 12%. Yield ≥ 40% of the 10-year G-sec (2.8% at 7%, about twice NIFTY's yield). Pass --gsec with today's yield. Dividends are taxed at your slab rate, with 10% TDS once a company pays you more than ₹10,000 in a financial year (s.194, from FY 2025-26). In the 30% slab a 3% yield nets about 2.1% (indian-tax-for-traders).
  • dividend-growth-pullback: yield ≥ 1.5%, dividend growth ≥ 10%, profit growth ≥ 12%, payout ≤ 60%, ROE ≥ 15%, D/E ≤ 0.5, and RSI(14) ≤ 40 from --prices-dir (daily CSVs named after the symbol or company, read with ../technical-analysis/scripts/ta.py) or an RSI column. Without either, the pullback rule is reported as not checked.
  • garp: trailing PEG between 0 and 1.5 (P/E ÷ 3-year profit growth if no PEG column), profit growth ≥ 15%, sales growth ≥ 10%, ROCE ≥ 18%, ROE ≥ 15%, P/E ≤ 40.
  • quality-compounder: ROCE ≥ 20%, ROE ≥ 18%, D/E ≤ 0.3, sales growth ≥ 10%, profit growth ≥ 12% over both 3 and 5 years, promoter ≥ 40%, pledge ≤ 1%, positive free cash flow. Board-run companies with no promoter fail the promoter rule by design.

Score: each rule scores 50 at its threshold and 100 at its "best" value (linear, capped), and the composite is the weighted average. Weights are in presets. Near misses failed exactly one rule and show which, with the value.

Filters and Screener.in queries

--where takes FIELD OP NUMBER only (> >= < <= == !=; and joins several). It is parsed by a regular expression and never evaluated, so __import__(...), function calls and field-vs-field comparisons are rejected. query prints the same rules as Screener.in query text, e.g. Return on capital employed > 20 AND Debt to equity < 0.3 AND Pledged percentage < 1. Query names seen in public Screener.in screens are marked verified (fields). Return over 1year was not found and is flagged unverified. Dividend growth has no known field and is left out. ≥ is written as >.

Hand-offs

  1. Read each passer's filings, governance and red flags with fundamental-screening. Check results timing with results-season.
  2. Time entries with breakout-planner (base and pivot) or momentum-scanner (RS and bursts).
  3. Before buying, check fit and concentration with portfolio-review.

Limits: the file is a snapshot from the export date, and stale columns give stale answers. RS is relative to your file only. Thresholds are judgement calls you can override. A screen cannot see related-party deals, auditor changes or surveillance status.

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.