domains / families / general / lemmaly

lemmaly

authentic

Algorithm-first discipline: state Big-O, data structure, and algorithm family BEFORE writing loops, queries, or recursion. Catches O(n^2), N+1, and brute-force defaults.

When to use

• Writing, editing, or reviewing code that involves loops, collections, lookups, searches, joins, recursion, graphs, queries, or any computation over more than a handful of items.
• About to write a `for` inside a `for`, `.find` / `.includes` / `.indexOf` inside a loop, `await` inside `for` / `map` / `forEach` over independent items, or one query per item in a collection.
• Auditing a codebase / PR for known anti-patterns (await-in-loop, `.includes` inside `.filter`, string-concat in loop, `SELECT *`, N+1, etc.).
• Reviewing AI-generated code that "looks idiomatic" but might hide O(n²) or N+1.
• **lemmaly** = think first (prevention).
• **complexity-cuts** = clean up bad Big-O (correction).
• **invariant-guard** = prove it's correct (verification).
• **mathguard** = beat the classical floor (acceleration).

When not to use

• **Not a substitute for profiling.** lemmaly forces asymptotic reasoning, not measurement. For constant-factor wins, latency tails, or I/O bottlenecks you still need a profiler.
• **Reasoning gate, not a code generator.** This skill changes how the model thinks before writing; it does not auto-rewrite existing code (use `complexity-cuts` for that).
• **English-language enforcement.** The rule catalog and prompts are English-only.
• **n < ~10 is exempt.** The protocol explicitly accepts trivial collections and one-shot setup code; do not waste time stating complexity for `for i in range(3)`.
• **Cannot prevent intentional brute force.** If the author writes a one-line justification ("n ≤ 100 in practice; readability matters more"), brute force ships. The skill only requires the justification, not its absence.
• **CLI scanner is separate.** The 59 rules are enforced by `lemmaly scan` in the upstream repo, not by this SKILL.md alone.

Skill page

Source