dotfiles

Audits a C# application or project path for performance issues — both syntactic (sync I/O on calling thread, LINQ inefficiency, excessive allocations, missing cancellation, improper async/await, memory leaks, inefficient collections, blocking thread pool) and semantic/data-flow (per-row DB commits, per-call DbContext/options rebuild, missing keyed indexes, DbContext-pooling and transaction-retry hazards, idempotent re-syncs lacking skip-if-clean, double broadcasts, unbounded fan-out) discovered by tracing hot paths through their callees, base classes, DI registrations, and notification handlers — then delegates to the feature-planning skill to produce a self-contained, phased, agent-ready optimization plan. Does NOT execute optimizations. A single failing test in the plan's Phase 0 baseline gate is a hard stop.

Pre-configured wrapper around package-dependency-tree-upgrade for Laravel projects. Audits first-party and organization packages in composer.lock to find those that support the current Laravel version but not the next, creates missing SP Projects features under an upgrade initiative (existing or newly created), and records blocker relationships in the correct upgrade order.

Audits a Laravel repository for custom Livewire components and produces an agent-ready upgrade plan for migrating from Livewire 3 to Livewire 4. Discovers components, classifies required changes by breaking-change category, and writes a structured plan using the feature-planning foundation.

Interactively create or update the northstar.md product vision document for a repository. Gathers vision, users, scope, deployment model, guiding principles (with BLOCK/WARN checks), and sanctioned feature tracking via AskUserQuestion. Pulls existing features from SP Projects (via SP Projects MCP) when available, then writes northstar.md to the repo's planning directory.

Audits a set of packages in a dependency lockfile to find those that support the current framework version but not the next, cross-references against an SP Projects upgrade initiative for existing features, creates missing features, maps inter-package dependencies, and records features as blockers in the correct upgrade order.

Generate or update end-user documentation for a set of implemented features by comparing a directory of plan files against the code that was actually written. Extends the user-docs skill — uses plans to locate and scope what to document, uses the code as the source of truth for UI labels, workflows, and behavior. Invoke when a feature epic has been implemented and you want accurate user-facing docs that reflect reality, not intent.

Converts British English spelling to American English in a file or directory. Scans text files for common British variants and rewrites them in-place. Accepts a file path or directory path as $ARGUMENTS.

Interactively scaffold the canonical convention-doc surface for any repository — root AGENTS.md + README.md, docs/standards/policies.md, sibling standards docs (logging conventions, etc.), and per-project AGENTS.md + README.md for each sub-project/module. Stack-aware (.NET, Laravel app_modules, Node/TS, Python, Go, generic). Detects stack, pre-fills baseline policies, interviews the user via AskUserQuestion, drafts everything, then applies review lenses.

Interactively create a new feature plan for any repository. Gathers requirements via AskUserQuestion, drafts a plan following discovered project conventions and general design principles, then applies plan-review lenses to produce an agent-ready plan written to the repo's planning directory.

Analyze an implementation plan file against the current state of the codebase, detect drift between what the plan describes and what the code actually contains, then interactively reconcile the plan so it accurately reflects reality and the remaining work. Invoke when asked to resync, refresh, reconcile, or update a plan after code has changed.

Analyze a directory of implementation plan files as a unified group for ambiguity, contradictions, missing information, and cross-plan inconsistencies, then interactively fill gaps and write enriched content back to each respective plan file. Invoke when asked to review, improve, or stress-test a set of related feature plans (e.g. an epic split across multiple plan files) before handing them to agents.

Analyze an implementation plan file for ambiguity, contradictions, or missing information, then interactively fill gaps and write the enriched plan back to disk. Invoke when asked to review, improve, or stress-test an implementation plan before handing it to an agent.

Agent orchestrator that executes all sub-plans produced by plan-split. Reads the dependency graph from sub-plan files, then spawns one coding sub-agent at a time in dependency-respecting order until all sub-plans are done. Sequential execution only — never runs sub-agents concurrently. Invoke when asked to execute, run, or implement a set of split plans.

Split a single fleshed-out implementation plan into sequentially ordered sub-epic directories, each containing a self-contained plan.md file. Output is compatible with plan-review-group (which scans for **/plan.md). Extends plan-split — use when splitting an epic-level master plan into sub-epics that each warrant their own full implementation plan, rather than lightweight task files.

Split a single fleshed-out implementation plan into sequentially ordered sub-plan files written alongside it in the same directory, with dependency (blockers/blocks) headers on each sub-plan. Sub-plans are always executed one at a time by plan-execute, so splits should optimize for clean sequential handoff, not concurrency. Invoke when asked to break a plan into phases or stages.

C#/.NET code review extending the base code-review skill. Checks for breaking changes, code quality, test coverage, and ASP.NET Core / Entity Framework patterns. Outputs AI agent prompts by default; use `--full` for a complete actionable report with per-file grouping, two severity tiers, and inline diffs.

Audits an Avalonia C# application or project path for performance issues (UI thread blocking, excessive bindings, missing virtualization, memory leaks via event handlers, sync I/O on UI thread, inefficient rendering), then delegates to the feature-planning skill to produce a self-contained, phased, agent-ready optimization plan. Does NOT execute optimizations. A single failing test in the plan's Phase 0 baseline gate is a hard stop.

PHP/Laravel code review extending the base code-review skill. Checks for breaking changes, code quality, test coverage, and Laravel patterns. Outputs AI agent prompts by default; use `--full` for a complete actionable report with per-file grouping, two severity tiers, and inline diffs.

Language-agnostic code review comparing current branch to main. Checks for breaking changes, code quality, test coverage, and framework-specific patterns. Outputs AI agent prompts by default; use `--full` for a complete actionable report with per-file grouping, two severity tiers, and inline diffs. Use `--lang <ext>` to filter to a specific file extension.

Generate or update developer documentation for a set of implemented features by comparing a directory of plan files against the code that was actually written. Extends the dev-docs skill — uses plans to locate and scope what to document, uses the code as the source of truth. Invoke when a feature epic has been implemented and you want accurate docs that reflect reality, not intent.

Audits a Laravel application or composer package module path for performance issues (N+1 queries, unbounded gets, missing cache, sync observers, route closures, fat controllers), then delegates to the feature-planning skill to produce a self-contained, phased, agent-ready optimization plan. Does NOT execute optimizations. A single failing test in the plan's Phase 0 baseline gate is a hard stop.

Apply this skill when removing database-level foreign key constraints and unique constraints from a Laravel project, replacing them with application-level enforcement. Handles both plan files (direct edits) and application code (new drop migrations + model boot logic).

Converts a repository from npm (or yarn) to bun as the package manager. Audits all affected files, applies changes, runs bun install, and verifies the build still works. Does NOT change build tool configuration beyond what is required to invoke the existing build tooling via bun.

Detects user-level skills that do not apply to the current repository and disables them in the repo's .claude/settings.json via skillOverrides. Discovers skills dynamically at invocation, analyzes the repo's tech stack, reasons about relevance, and asks for confirmation before writing.

Audits a single skill file for token consumption and performance anti-patterns, presents prioritized recommendations, then asks the user whether to apply them. Never changes the skill's stated goal.

Developer documentation writer for software applications

Canonical file operation rules shared across all skills. Read this file at the start of any skill that performs file operations.

End user documentation writer for software applications

Apply this skill when implementing concurrency guards, mutex patterns, idempotency checks, or duplicate-operation prevention — such as preventing double-submissions, race conditions on model state transitions, or any situation where lockForUpdate() might otherwise be used.

Apply this skill when writing or reviewing any code that sets or switches tenant context — including command loops iterating over multi-tenant records, or any call to TenantContext::applyById or Context::forget(TenantContext::ID).