ai-coding-protocols

Activate when an Engineering Manager needs to shape a rough initiative into a clear, scoped, outcome-oriented brief before execution.

Activate when reviewing branches, commits, or pull requests using the emoji-driven review protocol.

Activate when managing conversation history in long sessions, choosing compression strategies, or preserving critical information during context truncation.

Activate when designing agent systems, debugging unexpected agent behavior, or optimizing context usage and attention budgets.

Activate when hitting context limits, experiencing quality degradation in long sessions, or needing to extend effective context capacity.

Activate when generating in-code comments or system documentation using the Diátaxis framework.

Activate when testing agent systems, measuring output quality, or designing evaluation frameworks for AI-assisted workflows.

Activate when managing large outputs, persisting plans across turns, or using the filesystem as a context layer for agent workflows.

Activate when writing, reviewing, or testing PHP code to enforce clean code, strict typing, and PHPUnit standards.

Initial project analysis after vendoring — detects stack, conventions, and generates a project-overview skill.

Activate when starting LLM-powered projects, evaluating task-model fit, designing pipelines, or estimating costs for agent-assisted development.

Activate when exploring a codebase, tracing dependencies, or planning changes using recursive reasoning.

Activate when improving code quality, reducing complexity, or aligning code with standards without changing behavior.

Meta-skill for creating new project-specific skills as the codebase evolves.

Activate when designing MCP tools, custom agent tools, or optimizing tool descriptions and parameters for agent consumption.

Activate when writing, reviewing, or testing TypeScript code to enforce clean code and TDD standards.

One-line activation trigger — when should an AI agent load this skill?