Ponytail

Ponytail — Laziness Ladder Audit

Purpose

Catch over-engineering: unnecessary abstractions, unused dependencies pulled in for one-liners, reinvented stdlib/native-platform features, dead code, and premature generalization. This is the review half of ponytail; the generation half (the ladder itself) is always active via USER_RULES.txt → STATIC PRINCIPLES, so every diff should already be reasonably lean — this skill is for the cases that slip through or for auditing code this agent did not just write.

When to activate

• User asks to "review this diff for over-engineering", "does this need to exist", "simplify this", "audit the repo for bloat/dead code".
• Explicit command: /ponytail_review (diff) — see commands/ponytail_review.md.
• Post-implementation self-check before marking a task done, when the diff touched more files/lines than the request implied.

The ladder (source of truth — also in USER_RULES.txt)

Before writing code, stop at the first rung that holds:

1. Does this need to exist? → no: skip it (YAGNI).
2. Already in this codebase? → reuse it, don't rewrite.
3. Stdlib does it? → use it.
4. Native platform/language feature? → use it.
5. Installed dependency already covers it? → use it.
6. Fits in one line? → one line.
7. Only then: the minimum code that works.

Never on the chopping block: trust-boundary validation, data-loss handling, security checks, accessibility. Lazy about the solution, never about reading the problem — trace the real flow before picking a rung.

Audit procedure

1. Scope
   • Diff mode: git diff (staged+unstaged) or the PR's changed files.
   • Repo mode: walk source directories, excluding generated/vendored paths (node_modules/, dist/, build/, .venv/, lockfiles).
2. Flag candidates
   • New dependency added for something stdlib/native/one-line already covers.
   • New class/module/interface with a single call site and no stated extension need.
   • Wrapper components/functions that just forward to the thing they wrap.
   • Config options, feature flags, or parameters with no caller passing non-default values.
   • Dead code: unused exports, unreachable branches, commented-out blocks.
   • Duplicated logic that could reuse an existing in-repo utility (check before proposing a new one — pairs with repo-grounding).
3. Do not flag
   • Validation, error handling, security, accessibility, or anything load-bearing for correctness even if it adds lines.
   • Abstractions with a real second caller or an explicit near-term requirement.
4. Report — a delete-list, not a rewrite:
   • Table: file:line, what it is, why it's excess, suggested replacement (often "delete" or "inline").
   • Do not apply deletions without confirmation unless invoked as part of /cleanup with explicit APPLY.

Debt ledger (optional, lightweight)

If the user defers a ponytail:-style shortcut comment (e.g. a native feature used today that will need a real component later at scale), note it in the PR description or a TODO with an owner/trigger condition — don't let "later" become "never" silently. Do not create a dedicated ledger file unless the user asks for one (matches DOC HYGIENE: no orphan tracking files).

Output contract

• Delete-list table (file:line, issue, suggested fix).
• One-line verdict: how many rungs of the ladder the diff/repo violates, if any.
• If invoked mid-implementation (self-check), apply the obvious deletions inline before calling the task done; if invoked as a standalone audit, propose only — apply after confirmation.