A portable collection of Agent Skills that help AI coding agents build trading strategies, automation bots, and backtesting scripts on the official FYERS Developer API v3 (Indian markets — NSE/BSE/MCX). They import cleanly into Claude Code, Cursor, and Open Claw.
They target the public developer API (https://api-t1.fyers.in + the fyers-apiv3
SDK + WebSocket feeds) — not the FIA chat-assistant proxy (fia.fyers.in).
Each skill is a self-contained folder under skills/. The install command (below) discovers
them all.
fyers-trading— build/backtest/order plumbing on the FYERS API v3 (the core skill).fyers-supercharge— Supercharge Mode: an iterative multi-agent optimization loop that debates a strategy, generates optimized variants, backtests them, and searches for measurably better versions across rounds (keeping the original as the baseline). Depends onfyers-trading; install both together.
skills/
fyers-trading/ # The core skill (start here)
SKILL.md # frontmatter + instructions
.env.example # FYERS_APP_ID / FYERS_SECRET_ID / FYERS_REDIRECT_URI
requirements.txt # Default deps for generated strategy/bot/backtest code
references/ # Loaded on demand (progressive disclosure)
setup.md # venv + install steps, package list, troubleshooting
auth.md # OAuth v3 flow, appIdHash, daily token, refresh
endpoints.md # Full path + payload + ENUM CODE catalog
symbols.md # Symbol format (eq/fut/opt, weekly vs monthly), masters
market-data.md # quotes / depth / history / option chain / status
orders.md # place/modify/cancel, GTT, smart orders, positions
websocket.md # data / order / TBT sockets
backtesting.md # candles -> DataFrame -> strategy -> costs/pitfalls
indicators.md # TA-Lib install + indicator wrappers reference
quantstats.md # QuantStats install + tear-sheet / risk-metrics reference
rate-limits.md # 10/s, 200/min, 100k/day; error codes; retry policy
scripts/
fyers_login.py # OAuth login + daily token cache (~/.fyers/token.json)
fyers_client.py # reusable REST client (dry-run orders, 429 backoff)
fyers_symbols.py # download/cache symbol masters; resolve name -> symbol
helper.py # token-free order utilities: lot validation, tick rounding, DTE
option_chain.py # option chain helpers: parse, ATM, PCR, straddle, max pain
example_strategy.py # data -> signal -> DRY-RUN order skeleton
indicators.py # TA-Lib indicator wrappers (SMA/EMA/RSI/MACD/BBANDS/ATR/...)
quantstats_report.py # QuantStats tear sheet + risk metrics from a returns series
trade_logger.py # append-only JSONL audit log (~/.fyers/trades.jsonl)
validate-skill.sh # lints SKILL.md before importing
fyers-supercharge/ # Supercharge Mode (iterative multi-agent optimization)
SKILL.md # orchestrator + council + optimization-loop instructions
references/
optimization-loop.md # the 6-phase round lifecycle + consensus rule
council.md # 12-seat roster: domains, mandates, veto powers, owned metrics
variants.md # variant contract, strategy interface, Variant Library taxonomy
metrics.md # Comparison Dashboard metric set + where each number comes from
checkpoints.md # the user checkpoint every 3 rounds
deliverables.md # final outputs + success criteria
overfitting.md # robustness / anti-overfitting guardrails
scripts/
scorecard.py # standardized per-variant metrics (reuses quantstats_report)
dashboard.py # rank baseline + variants, render comparison dashboard
evolution_log.py # append-only JSONL journal of rounds / variants / decisions
assets/ # fixtures / images
- Create an app at https://myapi.fyers.in/dashboard/ and note the app id, secret, and redirect URI.
- Create a venv and install deps:
python3 -m venv .venv && source .venv/bin/activate && pip install -r requirements.txt(seereferences/setup.mdfor the package list and install troubleshooting). - Ask your agent to build a strategy/bot/backtest. By default the agent drives setup and
development with you conversationally: it scaffolds
.env(keys present, values blank) in the project folder and waits for you to fill in the values yourself, then runs the login flow and verifies it against/profile— seereferences/auth.md. It then builds the strategy incrementally, running each piece for real and showing you the output as it goes (seeSKILL.md's "Default mode: conversational execution"). Once the strategy runs, it automatically backtests it and generates a QuantStats tear sheet (Sharpe/Sortino/drawdown + monthly heatmap) without you having to ask, then offers Supercharge Mode. If you'd rather just get code without the agent running anything, say so explicitly. - Prefer to do it yourself instead?
cp skills/fyers-trading/.env.example .env, fill in the values, runpython skills/fyers-trading/scripts/fyers_login.py(caches the daily token), then verify withpython skills/fyers-trading/scripts/fyers_client.py profile.
The skill enforces these in any code it generates:
- Secrets via environment variables only — never hardcoded or committed.
- Order placement is dry-run by default; live trading needs an explicit opt-in.
- Confirm before any live place/modify/cancel.
- Respect rate limits (10/s · 200/min · 100k/day; order ops ≤10/s).
- Use WebSocket for live ticks, not polling.
- Tokens expire daily — a 401 means re-login, not retry.
Recommended: npx skills add (the skills CLI)
npx skills add FyersDev/fyers-skills # discovers all skills, prompts you to pick
npx skills add FyersDev/fyers-skills -g # global (user-level)
npx skills add FyersDev/fyers-skills --all # all skills, all detected agents, no promptsThis walks the repo, finds every SKILL.md under skills/, installs the ones you pick
(fyers-trading, fyers-supercharge), and wires them into the agent directories it
detects. Useful flags: -a/--agent <agents> to target specific agents, -l/--list to
preview without installing, --copy to copy files instead of symlinking. List or remove
later with npx skills list / npx skills remove fyers-trading.
Supercharge Mode depends on the core skill.
fyers-superchargecalls intofyers-trading's scripts, so install both (--all, or pick both at the prompt).
Requires a public repo.
skills addfetches over public GitHub and has no private-repo/token support — ifFyersDev/fyers-skillsis private the command 404s. Make the repo public, or use the manual install below.
Each skill's folder name matches its name: (fyers-trading, fyers-supercharge).
Clone the repo, then copy each skill folder from skills/ into the host's skills directory:
git clone git@github.com:FyersDev/fyers-skills.git && cd fyers-skillsClaude Code
mkdir -p .claude/skills # per-project (or ~/.claude/skills for per-user)
cp -r skills/fyers-trading skills/fyers-supercharge .claude/skills/Invoke with /fyers-trading / /fyers-supercharge, or just describe a FYERS task.
Cursor
mkdir -p .cursor/skills
cp -r skills/fyers-trading skills/fyers-supercharge .cursor/skills/Reload the window after copying.
Open Claw
mkdir -p ~/.openclaw/skills
cp -r skills/fyers-trading skills/fyers-supercharge ~/.openclaw/skills/Skills-directory paths vary by host version. If a host can't find the skill, check its docs for the active skills directory and confirm the folder name matches
name:.
skills/fyers-trading/scripts/validate-skill.sh skills/fyers-trading/SKILL.md
skills/fyers-trading/scripts/validate-skill.sh skills/fyers-supercharge/SKILL.md
python -m py_compile skills/fyers-trading/scripts/*.py skills/fyers-supercharge/scripts/*.py- Built against FYERS API v3 (
api-t1.fyers.in). Span-margin and EDIS still use v2 (api.fyers.in/api/v2). - The exact access-token TTL isn't published;
fyers_login.py --checkverifies liveness by calling/profilerather than guessing an expiry. - The refresh-token flow may be discontinued by FYERS — daily re-login is the reliable path. Verify against the live docs (https://myapi.fyers.in/docsv3) if you depend on it.
- For WebSocket streaming, install the official SDK:
pip install fyers-apiv3.
MIT — see LICENSE.