AI Agent Workflow Collection

Proactive codebase audit that discovers code quality issues, security vulnerabilities, performance problems, and architectural debt using parallel analysis agents, then creates well-structured issues in PinkRooster from confirmed findings. Use this skill whenever the user wants to audit their code, find problems, scan for issues, do a code review of the whole project, or says things like "audit the codebase", "find issues", "scan for problems", "what's wrong with the code", "check for security issues", "find tech debt", "code health check", "quality scan", or "are there any bugs". Also triggers when the user asks to analyze code quality and wants actionable tracked issues created from the findings, not just a report.

Interactive feature brainstorming through guided discovery, web research, and codebase analysis. Helps users who have a vague idea ("I want to improve X", "what if we added something for Y") explore possibilities through Socratic dialogue before creating well-formed feature requests. Use this skill whenever the user wants to brainstorm, explore ideas, think through a feature concept, or says things like "brainstorm", "what could we do about...", "I have an idea", "let's think about...", "explore options for...", or "help me figure out what feature to build". Also use when the user has a rough concept but needs help shaping it into something concrete — this is the skill for turning fuzzy thinking into sharp feature requests.

Analyze codebase for dead code, unused imports, inconsistencies, and structural debt, then scaffold a work package with cleanup tasks. Never modifies code directly — changes flow through the normal implement/test/commit pipeline. Use when the user says "clean up the code", "remove dead code", "tidy up", "find unused code", "code hygiene", or "clean up imports".

Scan the codebase for common code smells using threshold-based detection: long methods, god classes, long parameter lists, deep nesting, magic numbers, data clumps, and more. Produces a prioritized overview and proposes PinkRooster issues for confirmed findings. Faster and more systematic than pm-audit — uses quantitative metrics rather than exploratory analysis. Use this skill whenever the user says "code smells", "find smells", "check for code smells", "method too long", "god class", "refactoring opportunities", "what needs refactoring", "code quality metrics", "complexity check", or wants a quick structural health scan of their code without a full audit.

Mark tasks, issues, or feature requests as completed and report cascading state changes. Automatically completes linked entities when a work package auto-completes.

Analyze the codebase from a product manager's perspective and suggest realistic, user-facing feature enhancements. Cross-references existing items to avoid duplicates. Creates selected suggestions as feature requests.

Identify and remove stale, cancelled, rejected, or superseded PinkRooster entities (issues, feature requests, work packages). Keeps the project board focused on active, relevant work. Use when the user says "housekeeping", "remove stale items", "clean the board", "delete cancelled items", "prune old entities", or "tidy up the project".

Implement a task, phase, or entire work package. Reads context, analyzes target files, implements code changes, runs tests, and updates task states. Auto-transitions linked entities on start and completion. Use when the user says "implement", "start working on", "execute", "build this", or references a task/phase/WP ID and wants code changes made.

Pick up the next highest-priority work item and start implementing it. Finds WPs, unscaffolded Issues, and FRs by priority. Auto-scaffolds Issues/FRs without work packages, then implements. Use --auto for fully autonomous execution until all work is done. Use when the user says "what's next", "pick up work", "start implementing", "work on the next thing", or just "next".

Plan new work by creating issues or feature requests from a natural language description. Researches the domain online, analyzes the codebase for context, and decomposes complex descriptions into multiple entities when sensible. Use when the user describes work needed, a bug to fix, a feature to add, or says things like "we need...", "there's a problem with...", "I want to add...", "plan out...", or "let's track...". Also triggers for vague requests that need clarification before becoming actionable work items.

Refine a feature request by analyzing its current state, rewriting the description with full detail, filling in missing fields (business value, acceptance summary, category, priority), and adding structured user stories.

Scaffold a complete work package with phases, tasks, and dependencies based on a feature description or linked issue/FR. Analyzes the codebase to produce realistic target files and implementation notes. Auto-transitions linked entities to planning states. Use when the user wants to break down work, plan implementation, create a WP, or says "scaffold", "break this down", "plan the implementation", or "create tasks for...".

Show project status dashboard with issue/FR/WP counts, active items, blocked items, and priority next actions. Use when the user asks about project status, progress, what's happening, what needs attention, or what to work on.

Review and prioritize open issues and feature requests. Analyzes severity, age, and codebase impact to recommend priority adjustments and next steps.

Diagnose the root cause of a bug, error, crash, or unexpected behavior. Traces through code, logs, services, database state, and git history to find why something is broken. Researches error messages online for known issues. Produces a diagnosis with evidence and suggested fix. Use when the user reports a problem, error message, stack trace, test failure, or says things like "why is this happening", "this is broken", "I'm getting an error", "something's wrong with...", "debug this", or "figure out why...".

Verify acceptance criteria for a phase or entire work package. Runs verification based on each criterion's method (AutomatedTest, Manual, AgentReview) and records results via the MCP tool.