Whole-repo audit for over-engineering. Like ponytail-review, but scans the entire codebase instead of a diff: writes a ranked checklist to plans/AUDIT.md of what to delete, simplify, or replace with stdlib/native equivalents. Use when the user says "audit this codebase", "audit for over-engineering", "what can I delete from this repo", "find bloat", or calls ponytail-audit. One-shot report, does not apply fixes.

2026-06-252

ponytail-fix

really-knows-ai/gideas

Fix ponytail audit items from plans/AUDIT.md one at a time. For each item: implementer fix → ponytail-review → normal review → quality gate → mark done → commit. Cycle until reviewer is happy at each stage. Complements ponytail-audit.

2026-06-252

ponytail-review

really-knows-ai/gideas

Code review focused exclusively on over-engineering. Finds what to delete: reinvented standard library, unneeded dependencies, speculative abstractions, dead flexibility. One line per finding: location, what to cut, what replaces it. Use when the user says "review for over-engineering", "what can we delete", "is this over-engineered", "simplify review", or calls ponytail-review. Complements the correctness-focused reviewer subagent — this one only hunts complexity.

2026-06-242

execute-phased-plan

really-knows-ai/gideas

Use when executing a project folder under plans/ that contains SPEC.md, PLAN.md, and PHASE_XX.md files.

2026-06-242

fix-review-item

really-knows-ai/gideas

Fix the first incomplete checklist item in plans/<project>/REVIEW.md, commit the fix, then stop. Does not commit REVIEW.md — it is gitignored by design.

2026-06-232

implementation-review

really-knows-ai/gideas

Review the current repository state against a project's SPEC.md in plans/<project-name>/ and produce a checklist of issues in REVIEW.md. Use when you need a methodical spec-compliance audit of the present codebase.

2026-06-232

name	spec-lint-fix
description	Run markdown linting from tools/spec-lint/, fix issues, and rerun until clean

Overview

This skill enforces markdown quality after repository changes.

It runs the repository linter from tools/spec-lint/, fixes reported issues in markdown files, and repeats until bash lint.sh exits clean.

Entry Point

Read AGENTS.md at the repository root.
Run bash lint.sh in tools/spec-lint/.
If lint passes, report success and stop.
If lint fails, collect every reported issue (file, line, rule, and message) before editing.

Fix Loop

Repeat the following loop until lint passes:

Open each reported markdown file and apply minimal edits to satisfy lint rules.
Preserve technical meaning while fixing style and formatting issues.
Avoid broad rewrites unless required for a lint rule.
Re-run bash lint.sh in tools/spec-lint/.
Continue until the linter reports zero issues.

Rules

Run lint from tools/spec-lint/ only.
Use bash lint.sh exactly (do not substitute another command).
Fix only lint violations; do not introduce unrelated content changes.
Keep terminology and key decisions in AGENTS.md intact.
If a lint message is ambiguous, inspect the affected rule and apply the smallest valid fix.
Stop only when lint is clean and report the files changed.