Definitions for the terms the framework uses without re-explaining each time. One paragraph per term, ordered for reading top to bottom.
A plugins/<name>/ directory installable from this marketplace into Claude Code. Each plugin owns one domain (context, dev, vcs, pm, orchestrator, refine), ships its own README, CATALOG, and skills, and may add any Claude Code surface: agents, commands, hooks, rules, and MCP servers (.mcp.json). Plugins version independently via release-please.
The Git repository that publishes plugins. When you run /plugin marketplace add <owner>/<repo> Claude Code reads .claude-plugin/marketplace.json and offers the listed plugins for install. This repo (aidd-framework) is one such marketplace and versions independently of the plugins it ships.
Per-project context files under aidd_docs/memory/ (architecture, conventions, decisions, and similar) that give the AI durable knowledge of your codebase. Built or refreshed by aidd-context (the onboard / project-memory skills) and read by skills that need project context. "Onboard builds the memory bank" in the Quick start refers to this.
A self-contained workflow under plugins/<plugin>/skills/<NN-name>/. Triggered by a user phrase, a slash command, or an explicit Use skill <id> invocation. A skill owns a SKILL.md router, one or more atomic actions, and optional assets/ and references/. The SKILL.md name: is the folder slug (00-onboard); the invocation id is <plugin>:<folder>, for example aidd-context:00-onboard.
The shape every skill in this framework follows. SKILL.md is a pure router: it lists triggers, declares ## Available actions, and dispatches to one of them. It carries no business logic. The logic lives in the action files. This split keeps triggering predictable and keeps each step independently testable.
A single atomic step inside a skill, stored at plugins/<plugin>/skills/<NN-name>/actions/NN-name.md. Each action file contains only ## Inputs, ## Outputs, ## Process, ## Test, and optionally ## Depends on. Tests must be observable: a command to run, an artifact to check, or a side effect to verify.
A specialized AI persona under plugins/<plugin>/agents/<name>.md. Agents are scoped (for example executor, checker) and spawned by an orchestrator when a step needs isolation in a dedicated role rather than the main thread. Currently agents ship only in aidd-dev.
A coding standard the AI loads automatically on relevant files. Rules live under each tool's rules directory (Claude Code uses .claude/rules/) and can be scoped via paths: glob frontmatter. Rules without paths apply to all files. Rules describe how to write code; skills describe what to do.
A program declared in plugins/<plugin>/hooks/hooks.json that Claude Code runs at lifecycle events (pre-commit, post-tool, etc.). Hooks are how a plugin triggers deterministic side effects rather than asking the model to remember.
The short identifier in each plugin's Skills table, for example [2.1] for aidd-dev:01-plan. The first digit is the plugin ordinal; the second is the skill position inside the plugin. Bracket IDs are the stable references the per-plugin Skills tables use to point at a skill without spelling out the full id.
../README.mdfor the marketplace overview, plugin catalog, and install flow.ARCHITECTURE.mdfor the composition model and skill router pattern.