DietrichGebert/ponytail
52.7
Adequate · 13 November 2026
4k
lines of production code
Python
with JavaScript, TypeScript
7
measurements over time
What this system is
Features
Introduce Ponytail extension with mode management and skill shortcuts
Adds the pi-extension, which registers a /ponytail command for switching between runtime modes (lite, full, ultra) and setting defaults, alongside shortcut commands for skills like review, audit, gain, and debt. The extension includes argument autocompletion, a status bar indicator (with theme-aware styling and hiding options), and logic to parse commands, persist modes, and handle deactivation via specific phrases.
pi-extension · high confidence
Introduce agentic benchmark for multi-file coding sessions
Adds a new agentic benchmark suite under benchmarks/agentic that evaluates coding agents in real headless Claude Code sessions against a seeded codebase. The suite measures correctness, safety (adversarial input), source lines of code, and over-engineering, and includes new LLM-judged passes for completeness (complete.py) and answer quality (answer.py) to ensure agents actually deliver requested features and communicate clearly. It also adds a paired arm comparison tool (compare.py) for statistical analysis of benchmark results.
benchmarks/agentic · high confidence
Introduce the task board visualization
Added a self-contained HTML task board (\hooks/band/board.ts\) that renders project statistics, cost/time charts, and task lists directly from CSV ledger data without requiring a server or network. The board supports dark/light themes, responsive layouts, and interactive filtering, providing users with a visual summary of their agent's activity and costs. Also in this area: 4 commits (4 fixes) fixing hooks.
hooks/band · medium confidence · unverified
Introduction of Ponytail rule for efficient, minimal code generation
A new 'Ponytail' rule has been added to the Cursor configuration, establishing a 'lazy senior dev mode' that prioritizes the simplest working solution. This rule instructs the AI to avoid unnecessary abstractions, new dependencies, and boilerplate by reusing existing code, standard libraries, or platform features. It enforces a specific decision ladder (YAGNI, reuse, stdlib, etc.) before writing code, mandates fixing root causes rather than symptoms, and requires marking deliberate simplifications with 'ponytail:' comments. The rule also emphasizes that 'lazy' does not mean careless, requiring thorough problem understanding and leaving a small runnable check for non-trivial logic.
.cursor · high confidence
New benchmark arms for caveman and ponytail communication styles
Added new benchmark arms in the \benchmarks/arms\ directory to evaluate specific communication styles. The 'caveman' arm (\caveman.js\ and \caveman-SKILL.md\) implements an ultra-compressed communication mode that reduces token usage by dropping articles and filler words while maintaining technical accuracy, supporting various intensity levels. The 'ponytail' arm (\ponytail.js\) uses the repository's own \SKILL.md\ as the system prompt, serving as a baseline for the standard assistant behavior. A 'baseline' arm (\baseline.js\) was also added, which sends only the user task without any system prompt.
benchmarks/arms · high confidence
New benchmark harness with behavioral and correctness gates
The benchmarks directory now includes a comprehensive evaluation suite for Ponytail. It adds behavioral gates (behavior.js/behavior.yaml) to verify that the ruleset enforces specific engineering practices like leaving calibration knobs, providing requested explanations, and including runnable checks. It also introduces a correctness assertion (correctness.js) that executes generated code for email, debounce, and CSV tasks to ensure functional integrity, alongside existing line-of-code metrics. The suite supports multiple models (Anthropic, OpenAI, Gemini) via dedicated promptfoo configurations and requires Node.js ≥ 22.22.0 and Python 3.
benchmarks · high confidence
New hooks and runtime support for Claude, Copilot, Codex, Cursor, Qoder, and CodeBuddy
This change introduces the core Ponytail hook infrastructure, adding JSON configuration files for Claude, Copilot, Cursor, and Qoder, alongside JavaScript runtime scripts that manage mode activation, statusline nudges, and ruleset injection. The hooks now support multiple AI coding assistants (Claude Code, VS Code Copilot, Codex, Cursor, Qoder, and CodeBuddy) with host-specific logic for detecting the environment, writing mode flags, and injecting the appropriate ruleset into the model's context. It also includes a codebase map builder to help the model reuse existing code and shell scripts to display the active Ponytail mode in the statusline.
hooks · high confidence
New scripts for version consistency, rule-copy drift checks, and OpenClaw skill publishing
The scripts directory now includes build-openclaw-skills.js and publish-openclaw-skills.js to generate and publish OpenClaw/ClawHub skill packages with short descriptions, check-rule-copies.js to verify that host-specific rule copies match AGENTS.md and contain required invariants, check-versions.js to enforce a single pinned version across all plugin manifests and release tags, cursor-hooks.js to safely install and uninstall Cursor hooks without affecting other entries, and uninstall.js to clean up mode flags, statusline entries, and Cursor hooks left behind by host uninstallers.
scripts · high confidence
Ponytail 5.3.2 initial release with multi-agent support and lifecycle hooks
This release introduces Ponytail 5.3.2, a 'lazy senior dev' mode that enforces minimal, efficient code generation. It ships with native plugin support for a wide range of AI agents, including Claude Code, Codex, OpenCode, Gemini CLI, Qoder, Cursor, and Hermes Agent, utilizing lifecycle hooks to inject rulesets and register slash commands like \/ponytail-review\ and \/ponytail-audit\. The package is available via npm as \@dietrichgebert/ponytail\ and includes a comprehensive \INSTALL.md\ for setup across different environments. New features include a 'band' and 'board' for tracking task costs and agent time in Claude Code, and a rebuilt ruleset that prioritizes reusing existing code and standard libraries over adding new dependencies.
(repo-wide) · high confidence
Fixes
Ponytail plugin ported to OpenCode 2 API with improved command handling
The Ponytail plugin has been updated to support the OpenCode 2 plugin API while maintaining backward compatibility with V1. This change introduces a new plugin entry point in .opencode/plugins/index.js and separates the frontmatter parsing logic into ponytail-frontmatter.cjs to prevent loader conflicts. Key behavioral improvements include ensuring command arguments are preserved even when templates lack a $ARGUMENTS placeholder, correctly registering skills via the config hook for discovery, and fixing the /ponytail command to properly toggle modes or report the current level without saving 'review' as a persistent state.
.opencode/plugins · high confidence
Test coverage
Added comprehensive test coverage for Ponytail's core logic and host adapters; Added tests for the Ponytail extension's core logic and helpers.
Dependencies
Publish @dietrichgebert/ponytail as an installable npm package
The ponytail project is now available as a public npm package (@dietrichgebert/ponytail v5.1.0), making it installable for OpenCode, Pi, and Qoder. The package includes the main plugin entry point, hooks, skills, and a dedicated Pi extension, with scripts for uninstallation and Cursor hooks integration included in the distribution.
(dependencies) · high confidence
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
Score
- CAI 52 → 53 (+0.7)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.3) — scores are not directly comparable.
Lenses
- Code Health 42 → 42 (-0.2)
- Architecture 100 → 100 (+0.0)
- Maturity 63 → 65 (+2.3)
- Readiness 42 → 45 (+2.8)
- Security 85 → 80 (-4.9)
- Performance 100 → 100 (+0.0)
Resolved (3)
- Documentation: written for insiders (docs/cursor-hooks.md)
- Scanner failed to run — not a clean result
- Scanner failed to run — not a clean result
Changes since last survey
- 6 commits — 2 feature/other, 4 fixes
By area
- hooks/band — 4 commits
- (root) — 2 commits
Notable commits
- fix: fix(band): drop the duplicate? hint, count only project files, tighter details view (#1305)
- fix: fix(band): find the transcript without a classic hook (#1242)
- fix: fix(band): size the row to bodyColumns only in the terminal (#1302)
- fix: fix(band): watch only /clear, /resume and /branch, so other plugins' output is not labelled ponytail (#1291)
- change: chore: release v5.3.3 (#1307)
- change: docs: add PhotoReunion to Already built with Ponytail (#1313)
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
DietrichGebert/ponytail was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 13 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit 96608e078b2d5ea607c5c3b8b5a54d1c1fba6846 — the exact code this score is about.
- Scored under rubric-2026.11.3 — the same rubric and the same method as every other entry in this index.
- Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.61+preprod-f3449e6eb431.