copilot-developer

Lightweight bug fix handler. Creates a bug document with Five Whys root cause analysis, fixes the issue, writes regression tests, and creates a PR. Activates when the user says "bugfix", "fix this bug", "something is broken", or reports a defect that needs a quick fix.

Guides when and how to update domain skills, documentation files, and agents after code changes. Activates when the doc-sync-hint hook flags related skills/docs, when directly editing skill, doc, or agent files, or when the user requests documentation updates.

Execute an existing plan step-by-step on a dedicated git branch, with fast ralph loops per step and a deep final ralph loop that includes self-review and adversarial testing. Writes a timestamped journal with commit hashes and loop results. Supports both new multi-file story format (plan.md + design.md + prd.md) and old monolithic iteration format. Use when the user wants to implement a plan, says "execute the plan", "start building", "implement this", or references a plan.md file to work from.

Discover, catalog, and manage application features in docs/features/. Use when the user wants to map out what their application does, add a new feature to the catalog, review existing features, or understand the feature landscape before planning work. Auto-invoke when the user says "what features do we have", "add a feature", "feature catalog", "document this feature", or "map out the app".

Handles PR review feedback cycles. Reads PR comments, creates feedback.md as an immutable review record, applies fixes, and updates the PR. Activates when the user says "feedback", "PR feedback", "address review comments", "handle review", or has PR feedback to process.

Interactive planner that walks through a 7-phase discovery process producing altitude-layered documents: brief.md (WHY), prd.md (WHAT), design.md (HOW), plan.md (WHEN/ORDER). Supports subcommands for individual layers and GitHub sub-issue creation. Plans are scoped to a feature and saved in story directories. Auto-invoke when the user says "let's plan", "PRD", "before we code", "plan this out", "walk me through the requirements", or describes a significant change they want to build.

DEPRECATED: Use /research context instead. This skill is kept for backward compatibility only. Activates ONLY when the user explicitly says "/research-brief". For all other research triggers, use the /research skill.

Research and discovery skill with three modes: story mode (creates story directory, writes brief.md with RICE scoring, creates GitHub ticket), context mode (scans codebase, produces research brief for external AI tools), and spike mode (creates time-boxed investigation document). Activates when the user says "research story", "research spike", "research brief", "research context", "/research", "I want to research", or wants to investigate before planning.

Defines conventions for building AI agents with the Laravel AI SDK and Ollama. Activates when creating or modifying agent classes in app/Ai/Agents/, working with AgentResolver routing, agent testing with ::fake()/::assertPrompted(), streaming SSE responses, conversation memory, or when the user mentions AI agent, agent pattern, Laravel AI SDK, agent resolver, temperature, or apply blocks.

Builds production-grade Claude Code skills through interactive discovery. Guides through qualifying the idea, scoping, designing architecture, writing SKILL.md and supporting files, and integrating with boost.json and hooks. Activates when the user wants to create a new skill, says "build a skill", "new skill", "create a skill", "skill for", or describes automation they want Claude to learn.