Generate a cross-tool usage insights report from locally stored AI coding agent sessions (Claude Code, Claude Cowork, GitHub Copilot for VS Code/CLI/JetBrains, Cursor, Codex, Kiro, Antigravity, OpenCode) over the last N days, default 30. Use when the user invokes /agent-insights, asks for insights, analytics, or a report on their agent sessions, asks how they're using their coding agents or what's working or causing friction, wants to compare their coding agent tools, or asks what their AI coding patterns look like, even when they don't say "report".
Create layered system architecture diagrams using HTML/CSS templates with color-coded tiers and grid layouts. Best for technology stacks, microservices topology, and multi-tier application design.
Expert code review with a senior engineer lens — covers architecture, SOLID principles, security vulnerabilities, performance, error handling, testing gaps, and API contract changes. Use this skill whenever the user asks to review code, check changes, audit a diff, look over a PR, find bugs in their work, or wants feedback on code quality. Also trigger when the user says things like 'review my changes', 'check this before I push', 'what do you think of this code', 'any issues with this?', 'code review', 'look this over', or 'is this ready to merge'. Works with any programming language and any size of change.
De-slop pass for any text: mechanically scans for the statistical tells of AI writing (the "not X but Y" reflex, puffery, uniform cadence, …) and rewrites by meaning into the target register, tweet to academic paper. Use when the user says "declawed"/"deslop", asks to remove AI tells / humanize text / make it not sound like AI, names a specific tell to strip, or before publishing any agent-drafted prose.
Use when producing or updating a gap-analysis HTML report. Triggers on "gap analysis", "gap register", "gap index", "analyse the requirement docs", "find gaps between BPMN and requirements", or when the user references requirement / BPMN / FeatureSpec / RFP docs and asks for a structured comparison. Also triggers when the user points at a folder of existing gap-analysis reports and asks to refresh, re-roll up, or rebuild the master register/index. Produces a self-contained, theme-switching HTML report in the gap-analysis house style. Use this skill instead of free-form Markdown when the user wants a sharable HTML deliverable; do not use for prose summaries or open-question lists.
Complete Research → Plan → Implement workflow for complex tasks. Use this skill whenever the user faces: complex features spanning multiple files, bug investigations requiring root cause analysis, refactoring with architectural implications, or any non-trivial work where understanding the codebase first will prevent wasted effort. Also use when the user says things like 'help me figure out how to...', 'I need to understand before building...', or 'let’s research this before coding'. Skip for simple bug fixes, typos, or single-file changes.
Guide users through spec-driven development — from goal clarification through requirements, design, task planning, and incremental execution. Use this skill whenever a user wants to build a new feature with upfront planning, asks for a structured development workflow, mentions "spec-driven", wants to create requirements or a design doc before coding, or invokes /spec-coding. Also trigger when a user has existing spec files (requirements.md, design.md, tasks.md) and wants to continue or execute tasks from them.
Compact a spec plan (or design doc) markdown file after the implementation has shipped: strip the transient how — code snippets, step-by-step walkthroughs, file-by-file diffs — that the codebase now owns, keep the why — design decisions, key file references, changelog. Use when the user wants to compact, shrink, or trim a plan/spec file post-implementation, says the plan is bloated or too long now that the feature shipped, or invokes `/spec-compact`. Renaming files, fixing dead links, and index updates belong to `spec-lint`.