v4.1.0: three more agents

v4.1.0: three more agents

Same lazy senior dev, three new places to put him. No change to the rules or the
benchmark numbers. This release is reach, not behavior.

New adapters

• Codex (#7): full plugin with a marketplace manifest, lifecycle hooks, and a
  shared runtime that keeps the Claude path byte-identical. Invoke as @ponytail,
  @ponytail-review, @ponytail-help.
• Pi (#1): a Pi extension with /ponytail mode control, per-session
  persistence, and system-prompt injection. Install with
  pi install git:github.com/DietrichGebert/ponytail.
• Kiro (#6): a drop-in steering file at .kiro/steering/ponytail.md.

Eight hosts now: Claude Code, Codex, Pi, Cursor, Windsurf, Cline, Copilot, Kiro.

Tooling and docs

• Rule-copy drift check (#3, #9): node scripts/check-rule-copies.js keeps
  every adapter copy aligned with AGENTS.md and guards the SKILL.md source against
  silently losing a rule.
• Agent portability doc (#2, #10): one table mapping each host to its files.
• Cross-platform fix for the hook compatibility test (Windows os.homedir).

Under the hood

The hooks were factored into a shared runtime (ponytail-runtime.js) and
instruction builder (ponytail-instructions.js), so Claude, Codex, and Pi load
the same rules instead of each duplicating the logic. Every adapter is
unit-tested; the Codex and Pi extension APIs were verified against their official
docs.

Full changelog: v4.0.0...v4.1.0

v4.0.0: production grade, still lazy

The hardening release. Three new reflexes, about ten lines of prompt:

• One runnable check. Non-trivial logic leaves behind the smallest test that fails if the logic breaks. No frameworks, no fixtures. One-liners stay test-free.
• Ceilings are named. A ponytail: shortcut with a known limit (global lock, O(n²) scan) must name the limit and the upgrade path in the comment.
• Robust beats flimsy at equal size. Between two same-size stdlib options, take the one that is correct on edge cases.

Benchmarked

Six tasks, three arms, same model, adversarial security and concurrency probes. Every arm passes every probe. Then the agreement ends:

	No skill	Caveman	Ponytail v4
Lines of code	3,629	1,440	490
Agent tokens	430,697	290,546	229,370
Surprise-extension lines	1,115	413	96

Full data and methodology: benchmarks/

Also

• All cross-agent rule files updated (Cursor, Windsurf, Cline, Copilot, AGENTS.md)
• ponytail-review no longer flags the minimal check as bloat
• README grew a chart

v1.0.0 — He ships.

The lazy senior dev, installable.

• YAGNI ladder, 3 intensity levels (lite / full / ultra)
• Claude Code plugin: auto-active sessions, [PONYTAIL] statusline badge, /ponytail-review, /ponytail-help
• Rules files for Cursor, Windsurf, Cline, Copilot, Aider
• Benchmarked: −16% tokens, ~4× faster, 293 → 47 lines of code (benchmarks)

Install

/plugin marketplace add DietrichGebert/ponytail
/plugin install ponytail@ponytail

The 246 lines nobody wrote have never caused an incident.