pi-package-goal

Resolve short approval against the latest checkpoint. Use for "lgtm", "looks good", "approved", or "go ahead"; not risky or review feedback.

Re-check a prior assistant claim. Use when user says "nack", "not convinced", or "check again"; not for review feedback.

UI/UX routing front door. Use as the front door for broad or ambiguous UI/UX work that needs the right build, redesign, audit, design-system, accessibility, or browser specialist. Do not use when the user already names a narrower UI skill.

Proposes UI Vault resource upgrades for a selected webpage. Use when the user asks for libraries, assets, effects, or inspiration for an existing page. Do not use for implementation or greenfield design.

Upload, list, or delete files with s3upload and return expiring Azure links. Use for configured temporary Azure storage; not for S3 or cloud setup.

Ship local Git changes by inspecting, polishing, validating, split-committing, and pushing. Use for commit, push, ship, delivery audit, or delivery blockers; not deploys or releases.

Remove proven unreachable code in small validated batches. Use when asked to find and delete dead, stale, orphaned, or unused code; not for dependency pruning, over-engineering reports, or speculative cleanup.

Redesigns a GitHub repository README or creates GitHub-safe README SVG assets from real project evidence. Use for README visual overhauls, heroes, section banners, diagrams, badges, or read-only README audits; not general website UI or ordinary documentation maintenance.

Run broad, open-ended or continuous roadmap-driven repository improvement as validated slices. Use when the agent should find and fix weaknesses, bugs, UI, tests, architecture, performance, or pipeline issues; not a single known defect or fixed plan.

Existing synthetic check. Use when validating the fixture.

Persistent ultra-compressed communication style without losing technical detail. Use for caveman mode, less tokens, be brief, or /caveman. Changes prose, not implementation scope.

Audit the whole repository for over-engineering, bloat, and deletion candidates. Use for audit this codebase, find bloat, or /ponytail-audit; use ponytail-review for a diff. One-shot report only; does not apply fixes.

List every `ponytail:` shortcut or deferral as a debt ledger. Use for ponytail debt, deferred shortcuts, or /ponytail-debt. One-shot report; changes nothing.

Show ponytail benchmark-median impact as a compact scoreboard. Use for ponytail gain, savings, impact, or /ponytail-gain. One-shot display, not a per-repo result.

Show the ponytail command and mode reference card. Use for ponytail help, available commands, or /ponytail-help. One-shot display, not a persistent mode.

Review a diff or scoped code change only for over-engineering and deletion opportunities. Use for simplify review, what can we delete, or /ponytail-review; use ponytail-audit for the whole repository. Does not review correctness.

Persistent YAGNI and shortest-safe-diff implementation mode. Use when the user says ponytail, be lazy, simplest/minimal solution, do less, or complains about over-engineering. Not a one-shot audit or review.

Run an available second-model review helper for closeout. Use when the user requests autoreview, Codex/Claude review, or pre-ship external review; not ordinary self-review.

Iterate on an existing PR, MR, or shelved changelist until Greptile reaches 5/5 with no unresolved comments. Use only when that review context, tooling, auth, and external-action approval exist.

Search for one unknown evidence-backed bug, then hand the reproduced defect to diagnosis. Use for bug hunts or unexplained failing validation; not an already-reported defect with a known repro.

Diagnose a specific reported failure or performance regression to root cause: reproduce, minimize, hypothesize, instrument, fix, and regression-test. Use for concrete broken, throwing, slow, or failing behavior; use bug-harvest when no defect is known yet.

Extract proven duplicate code inside a bounded folder and expose divergence bugs. Use for evidence-backed reuse or deduplication; not a topology-only folder split or speculative abstraction.

Refactor one folder's topology into coherent subfolders, with shared code only when duplication is proven. Use for a named directory split or organization task; not repository-wide architecture.

Produce an evidence-backed repository or folder audit report and prioritized plan. Use for health, risk, quality, security, testing, architecture, or modernization assessment; not implementation unless the user separately asks to apply findings.

Anti-slop taste rules for marketing sites, landing pages, portfolios, and editorial redesigns. Use when visual distinctiveness is the main goal; not product dashboards, data tables, or multi-step application UI.

Build distinctive, production-grade working web components, pages, or product interfaces. Use for greenfield frontend implementation. Do not use for image-only concepts, audit-only requests, or an existing-site redesign that must preserve structure.

Codex/GPT-only supporting rules for landing pages, GSAP motion, AIDA, and anti-slop polish. Use when that model-specific constraint is requested; not a general frontend router.

Structural anti-AI-slop design audit, redesign, or study from a page, URL, or screenshot. Use when the user explicitly invokes Hallmark or asks for its audit, redesign, study, or design-DNA workflow; use frontend-design for ordinary greenfield implementation.

Premium styling companion for an existing UI workflow. Use when the requested direction is soft, spacious, premium, luxury, or agency-polished.

Apply only to a named brutalist direction. Use for Swiss-grid, terminal, industrial, raw, or mechanical visual language after choosing the main build/redesign workflow.

Minimalist style overlay for a selected UI workflow. Use when the user explicitly requests minimalist, editorial, warm-monochrome, or Notion/Linear visual language.

Audit and upgrade an existing website or app UI while preserving functionality and implementation boundaries. Use for targeted visual modernization, not greenfield builds or audit-only reports.

Deep web/mobile design-system and accessibility guidance covering color, typography, layout, motion, dashboards, and interaction quality. Use for design decisions or review, not primary implementation when a build/redesign skill fits.

Create new agent skills with proper structure, progressive disclosure, and bundled resources. Use when user wants to create, write, or build a new skill.

Research with rforge provenance. Use for literature search, OSS study, systematic review, evidence extraction, meta-analysis, or review packages.

Rank noisy folders for skill-folder-refactor. Use when asked to find refactor candidates, noisy folders, or top folders/subfolders to split.

Maintain source-backed project docs and LLM wiki knowledge bases. Use for README, architecture/onboarding docs, docs drift, project wiki/entity pages, raw-source-to-wiki compilation, or wiki linting. Do not use for prose polishing, literature search, or code graph generation.

Use when generating brand-kit image prompts, logo-system boards, visual identity decks, or high-end brand guidelines.

Use only when a project needs the original v1 taste-skill behavior instead of the default v2 design-taste-frontend skill.

Use when a model must produce complete unabridged code or assets and avoid placeholders or truncation.