/improve-lint

Run the AI-augmented lint self-improvement loop on the current project. Scans `.lint-history/` for recurring violation patterns (≥3 hits in 30d window), drafts a Claude-ready prompt to author a new semgrep rule for the top candidate, and surfaces the proposal under

Boost this command

heymegabyte@heymegabyte

GitHub

From plugin

heymegabyte-claude-skills

2153 skills27 agents53 commands

Install

> /plugin marketplace add heymegabyte/claude-skills

How it fires

How this command gets triggered: by you, by Claude, or both.

Fires itselfClaude auto-loads it when your prompt matches the work.
You can call itInvoke it directly when you want it.
Slash command/improve-lint

Context preview

What this command does when you run it.

Run the AI-augmented lint self-improvement loop on the current project. Scans `.lint-history/` for recurring violation patterns (≥3 hits in 30d window), drafts a Claude-ready prompt to author a new semgrep rule for the top candidate, and surfaces the proposal under

Command definition

improve-lint.md

name: improve-lint
description: Run the AI-augmented lint self-improvement loop on the current project. Scans `.lint-history/` for recurring violation patterns (≥3 hits in 30d window), drafts a Claude-ready prompt to author a new semgrep rule for the top candidate, and surfaces the proposal under `.lint-history/proposals/<ts>.md`. Non-blocking analysis. See rules/lint-doctrine.md § Self-improving.
argument-hint: "[project-dir]"
user-invocable: true

/improve-lint

Closes the self-improvement loop on the project's lint output. Captures patterns that recur, surfaces them as draft semgrep rule candidates for codification.

What it does

1. Scans `.lint-history/*.log` from the last 30 days 2. Clusters violations by rule-id (ESLint / sonarjs / semgrep / unicorn / @typescript-eslint / etc.) 3. For patterns with ≥3 hits in the window, writes a proposal to `.lint-history/proposals/proposal-<ts>.md` 4. The proposal contains a Claude-ready prompt that drafts the YAML rule + cross-link narrative

Output shape

▸ Scanning lint history (last 30d)
  ✓ N log file(s) in window
▸ Clustering violations by rule-id
  · 6x unicorn/prefer-node-protocol
  · 6x @typescript-eslint/no-explicit-any
▸ Drafting proposal(s)
  ✓ Proposal written: .lint-history/proposals/proposal-<ts>.md

After proposal lands

1. Read the proposal 2. Paste the Claude-ready prompt into a new conversation (or invoke `claude` inline) 3. AI drafts the semgrep YAML at `~/.agentskills/templates/lint-stack/semgrep-custom/<topic>.yml` 4. Cross-link from owning domain rule (`code-style.md`, `rxjs-first-angular.md`, etc.) 5. Append a row to `rules/lint-doctrine.md` § "Codified incidents" 6. Commit + push agentskills (auto per `main-only-branch` + `brian-preferences` git policy) 7. Distribute: next `install-lint-stack.sh` pulls the new rule into every project

Invocation

bash ~/.agentskills/bin/lint-auto-improve.sh             # current dir
bash ~/.agentskills/bin/lint-auto-improve.sh /path/to/X  # named dir
# OR via npm script (added by install-lint-stack)
npm run lint:improve