agentleFS
ShubhamkumartrichyShubhamkumartrichy/claude-trading-skills-india/skills/thesis-tracker/SKILL.md

thesis-tracker

Skill64 starsChanged 16 days agoMITView on GitHub

Keeps every Indian stock or F&O trade idea as a written, falsifiable thesis in one JSON file the trader owns, and walks it through an enforced lifecycle (idea, watching, entry-ready, active, closed, or invalidated / expired) with a dated history. Flags reviews due, kill levels hit from a price file, results and expiry dates ahead, and delivery holdings about to turn long-term (STCG 20% to LTCG 12.5%). Closing reviews split outcome from process (right/wrong thesis x good/bad result), stats by source skill and setup, export to the trading-discipline journal. thesis.py, stdlib only, offline.

7 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use cim-builder, morning-note and 8 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • AnthropicSkill

    cim-builder

    Structure and draft a Confidential Information Memorandum for sell-side M&A processes. Organizes company information into a professional, investor-ready document with consistent formatting and narrative flow. Use when preparing sell-side materials, drafting a CIM, or organizing company data for a sale process. Triggers on "CIM", "confidential information memorandum", "offering memorandum", "info memo", "draft CIM", or "sell-side materials".

    Also in 3 projects that use this

  • AnthropicSkill

    morning-note

    Draft concise morning meeting notes summarizing overnight developments, trade ideas, and key events for coverage stocks. Designed for the 7am morning meeting format — tight, opinionated, actionable. Triggers on "morning note", "morning meeting", "what happened overnight", "trade idea", "morning call prep", or "daily note".

    Also in 3 projects that use this

  • +8 more from anthropics

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

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

    Every command takes --json and --as-of YYYY-MM-DD (default today, IST).

    line 49
  2. Very specific:

    Never edit the past: Do not rewrite history entries or the JSON by hand.

    line 79

“Common” is measured: kept by at least 6 in 10 of the 7 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

Thesis Tracker

A trade without a written reason cannot be reviewed; it can only be remembered, and memory edits itself to protect the ego. This skill makes the reason, the entry trigger and the conditions that prove it wrong explicit before money moves, then holds the trader to them afterwards.

Commands

T="python3 scripts/thesis.py --store ~/trading/theses.json"   # the store lives with the user's files
$T add TCS --direction long --horizon swing --source breakout-planner --setup breakout --tag it \
   --statement "Deal wins revive IT; the 8-week base breaks out and holds above 3400" \
   --trigger "daily close above 3400 on 1.5x volume" --trigger-price 3400 \
   --kill "Q2 TCV below last quarter" --kill-price 3330 --stop 3340 --target 3520 --target 3650 \
   --risk-pct 1 --catalyst 2026-11-20:"Q2 results"
$T add NIFTY --direction short --horizon swing --segment options --expiry 2026-11-24 ...
$T transition T0001 entry-ready
$T transition T0001 active --entry-price 3405 --qty 100 [--date 2026-11-14]
$T update T0001 --stop 3390 --note "trailed below the swing low"
$T note T0001 "Held the 20-DMA on falling volume" --reviewed       # schedules the next review
$T due --prices ltp.csv [--days 7] [--ltcg-days 45]                 # start of every session
$T list [--state active] [--symbol TCS] [--tag it] [--due] [--all];  $T show T0001;  $T stale --days 21
$T transition T0001 closed --exit-price 3520 --reason target --charges 160
$T postmortem T0001                     # draft: facts from the record + the questions to ask
$T postmortem T0001 --thesis right --stop-honoured yes --mistake none --lesson "..."
$T stats [--since 2026-04-01]
$T export --journal thesis_journal.csv [--closed-only] [--overwrite]
$T hypothesis --symbol DIXON --expected-win-rate 45 --margin 10

Every command takes --json and --as-of YYYY-MM-DD (default today, IST). Bad input prints one error: line and exits 2. --prices is any CSV with symbol and ltp (or close) columns.

Lifecycle

FromAllowed toNeeds
ideawatching, entry-ready, invalidated, expiredstatement (at add)
watchingentry-ready, invalidated, expired
entry-readyactive, watching, invalidated, expiredtrigger, kill criteria, stop (premium for bought options)
activeclosed--entry-price --qty to enter; --exit-price --reason to close
expiredwatching (revive)
closed, invalidatednothing: finalinvalidating needs --note saying what proved it wrong

Illegal moves are refused with the allowed list. Every change lands in the thesis history with its date; update records old and new values. Widening a live stop, or rewriting the statement or kill criteria while active, is allowed but flagged in the history and counted against plan adherence. Closed theses are frozen (later thoughts go in note). Reviews default to 7 / 30 / 90 days by horizon.

Why written kill criteria matter

The stop limits the loss; the kill criteria end the idea. "Q2 deal wins below last quarter", "stock enters ASM stage 2", "promoter pledge rises" are thesis-breaking even when price is above the stop. Written down before entry, they turn a later exit from a negotiation with hope into a lookup. A thesis without them cannot reach entry-ready.

How Claude uses it across sessions

  • The store file is the memory. At the start of a session where the user trades, run due (with a prices file if live quotes are available) and lead with kill hits, then reviews due.
  • Never edit the past. Do not rewrite history entries or the JSON by hand; use update / note, which keep the earlier values. Never delete a thesis.
  • Ask before closing, invalidating or expiring. due and stale only suggest; the trader confirms the state change, the fill price and the reason.
  • Register ideas from other skills with --source (breakout-planner, results-season, the stock screener, momentum-scanner, pair-trading) so stats can say which sources actually make money.

Postmortem and stats

The postmortem separates the decision from the dice. Result comes from the record (₹ and R, using the stop at entry). Whether the thesis was right comes from the trader: did the written statement come true? The quadrant is right/good (earned), right/bad (variance or execution), wrong/good (luck: do not repeat it) or wrong/bad (deserved); zero P&L is breakeven, never a loss. Plan checks (entry within 1% of the trigger, exit not beyond the stop, early exit, stop widened) are filled from the record and can be overridden with --entry-at-trigger, --stop-honoured, --exited-early. stats gives win rate, average R, expectancy, results by source, setup, tag, horizon, direction and segment, the quadrant table, plan-adherence rate (and followed vs broken expectancy) and the most common mistakes. Fewer than 30 closed theses: every number is provisional.

India-specific notes

  • Results dates are the commonest catalyst: add them with --catalyst; due warns when one falls inside an active holding. Read the reaction with results-season.
  • LTCG clock. For long delivery positions, due uses ../indian-tax-for-traders/scripts/tax.py (listed equity is long-term after 12 months) to list holdings turning long-term within --ltcg-days, with the STCG-vs-LTCG saving on the current gain. It never overrides a kill level.
  • F&O expiry. Futures and options theses need --expiry; due lists it ahead and flags an expired contract. A bought option with no stop counts the whole premium as risk.
  • Surveillance. Entry into ASM/GSM, an F&O ban or a trade-for-trade move belongs in --kill: margins jump and exits thin. Shorts cannot be carried in delivery, so they need another segment.

How it connects

  • trading-discipline: run its pre-trade gate before transition ... active; export writes rows in its journal format (date,symbol,side,entry,exit,stop,qty,charges + exit_date,setup,mistake, segment,notes), so ../trading-discipline/scripts/discipline.py review and ../position-sizing/scripts/risk.py journal read them. Active theses export with a blank exit (open trade); use --closed-only for risk.py, which has no open-trade concept.
  • backtest-lab: hypothesis prints a card (statement, evidence for and against, the backtest to run with ../backtest-lab/scripts/backtest.py, pass marks set in advance, kill criteria, sample size). Claude fills it in conversation; only a surviving idea is registered with add.
  • portfolio-review checks the whole book, this skill why each position is in it; position-sizing sizes it.

Output

Thesis check as of Tue 20 Oct 2026 IST (1 open in /home/me/trading/theses.json)
Target reached (1)
  T0001 TCS  LTP ₹3,600 past ₹3,520  unrealised +₹19,500 (+3.00R): book, trail, or restate the thesis
Turning long-term within 30 days (1)
  T0001 TCS  bought 2025-11-03: wait 15 days (sell on or after 2026-11-04) to convert STCG 20% into LTCG 12.5%

Storage: one indented JSON file with schema_version; each save writes a temp file and renames it over the store, keeping the previous version as theses.json.bak. A store from a newer version is refused, as is a store that changed on disk mid-command. The script writes only the store (and the journal CSV you name); default ./theses.json in the current folder, never inside the skill.

Limits: no prices are fetched; worded kill conditions are checked by the trader, not the script. Holidays and lot-size changes are not modelled. Tax figures ignore cess and the ₹1.25 lakh exemption.

Educational discipline tool, not investment advice. The thesis and the decision stay yours.

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.