lint-story

Advise on style and trade setup issues in a Trade Story. Warnings only - nothing here stops a story becoming a strategy.

SKILL.md

GitHub

Lint a Trade Story

Follow ${CLAUDE_SKILL_DIR}/../../story-skills.md and ${CLAUDE_SKILL_DIR}/../../reading-stories.md.

Warnings, never errors. Nothing here blocks a story from becoming a strategy — say so if a trader treats a warning as a failure.

Read the story

Run:

node "${CLAUDE_SKILL_DIR}/../../../dsl/src/verify.js" <path>

for merged and hints. If errors is not empty, this story has structural problems that come before style: say so in one line, point the trader at /verify-story, and stop.

This story has errors to fix first. Run /verify-story, then I'll check the style once it's clean.

No errors: print nothing for this and carry on.

Read the indicators the story names

Where the story names an indicator, what that indicator means can settle a clause that would otherwise read as ambiguous.

Two places may hold such notes: the built-in ${CLAUDE_SKILL_DIR}/../../../extras/indicators/, and a folder given as this skill's second argument. Consult each one that exists: read its README.md and follow it — the README says how the folder is organised and how to resolve a named indicator against it. Read nothing else there.

Where neither settles it, fall back on what you already know — a common indicator needs no file — and look it up when your own knowledge falls short. Check your memory first, so a lookup already made is not repeated; and once a lookup has told you what an indicator means, save it to your memory.

Do all of this silently — print nothing for it.

Lint

Print Checking style, then apply every rule in ${CLAUDE_SKILL_DIR}/../../../dsl/lint-rules.md and ${CLAUDE_SKILL_DIR}/../../../dsl/lint-rules-interview.md to the merged clauses in merged. Those two files hold the rules between them. Never suggest editing the story to fix something the defaults supplied.

Read stories/.settings/lint-overrides.md if it exists. A rule listed there produces no output — not a warning, not a note that it was suppressed.

Collect every warning before reporting. Where a rule plainly does not apply, say nothing rather than forcing it to fit.

Report

The progress lines are printed as the work happens. Never print one a second time when reporting, and never print a line of your own alongside them.

Nothing to flag:

✅ Nothing to flag.

Each warning — the rule id, the line and its text where one line is responsible, then the rule's Warn wording verbatim:

Warning on line 5 (both-directions):
Then    Go Long or Short, respectively
This story trades both directions. Backtest results will blend two different setups.

Where no single line is responsible, drop the line and its text:

Warning (both-directions):
This story trades both directions. Backtest results will blend two different setups.

Close with one status line. The count is not the point, so word it as what was found — one, two, or more than two:

⚠️ Found a potential problem:
⚠️ Found a couple of potential issues:
⚠️ Uncovered some potential issues:

Then a single AskUserQuestion offering Elaborate, Override and No.

Elaborate starts from the rule's Because and applies it to this story: name the clauses at fault, say what it would cost this particular setup, and what to change. Repeating Because unchanged is not an answer.

Override appends the rule id to stories/.settings/lint-overrides.md, scoped to this story unless the trader says otherwise, and creates the file if it is absent.