Glossary
Purpose #
A single page that defines every load-bearing term used elsewhere in the wiki, with a link back to its authoritative definition.
Big Picture #
These terms come from .docs/docs/specs/, .docs/ARCHITECTURE.md, and the ADRs. Where a term is normative, the spec wins.
Terms #
Adapter #
A package implementing the Adapter contract for one specific AI coding CLI (Claude Code, Codex, Copilot). Emits an engine directory + agent rules file. Siblings with stack packs; never imports one. See concepts/adapters-and-stack-packs.md.
Adapter manifest #
The static identity of an adapter — name, target CLI version, declared conformance level, capability declarations, gaps. Lives at-rest in adapter.md (YAML frontmatter) and at-runtime as the manifest field on the exported adapter. Reconciled with @aidokit/core schemas per ADR-0012.
Agent rules file #
The auto-loaded project rules file the target CLI reads at session start. Claude Code: CLAUDE.md. Codex: AGENTS.md. Copilot: .github/copilot-instructions.md.
AidoError #
The error envelope from @aidokit/core. Carries a machine-readable code, human message, optional details, and hint. CLI maps code onto exit codes per .docs/docs/specs/cli-reference.md §10.2.
Brownfield #
A project that already has source code, docs, or PRDs at init time. aidokit init --brownfield ingests existing artifacts as source docs (CHANGELOG.md [0.5.0]).
Capability declaration #
An adapter manifest's honest disclosure of what it writes (writesPaths), runs (runsShellCommands), and contacts (networkCalls). Verified by the conformance harness; misdeclaration is a conformance violation. See .docs/docs/specs/security-model.md §8.
Capability profile #
The permission level a role runs under: read-only, docs-only-write, or scope-limited-write. Adapters map these to their CLI's mechanism (Claude Code hooks vs Codex sandbox). See .docs/docs/specs/adapter-contract.md §9.
Conformance level #
One of Minimum, Standard, Strict. Declared in the manifest, verified by the harness in @aidokit/core. Declared-greater-than-verified is a CI failure. See concepts/conformance-levels.md.
Conformance harness #
The runner exported by @aidokit/core as runAdapterConformance / runStackPackConformance. Executes the level's checks against the target. Required in every adapter/pack CI. See .docs/docs/specs/conformance-levels.md §6.
DetectContext #
The lighter context passed to StackPack.detect(). Sandbox-style: fileExists, readFile, glob, readManifest, manifestHasDep, os. No shell, no network. See .docs/docs/specs/stack-pack-contract.md §7.1.
Dogfood gate #
The byte-compare invariant: from Phase 8, @aidokit/adapter-claude-code must emit a .claude/ tree byte-identical to this repo's hand-built .claude/. CI-enforced. See concepts/dogfood-gate.md and .docs/ARCHITECTURE.md §20.
EmittedFile #
A { path, content, mode? } tuple returned by adapter emit methods. Path is relative, forward-slash, no ... Mode is 0o755 for executable scripts. See .docs/docs/specs/adapter-contract.md §6.1.
Engine directory #
The adapter-specific config directory under the user's project root: .claude/, .codex/, or .copilot/. Owned by the adapter; user customisation goes elsewhere.
Interpolation #
The {{varName}} substitution helper in @aidokit/core. The only template mechanism allowed — no Handlebars, EJS, etc. Variables are a small fixed set per package. See .docs/ARCHITECTURE.md §11.2.
Kit content (v4) #
The skills, agents, hooks, slash commands, output styles, and schemas that @aidokit/adapter-claude-code emits. Reference lives in .docs/context/claude-workflow-kit-v4-base-part-{1,2,3}.md. See concepts/kit-content-v4.md.
MCP catalog #
The TypeScript const array in @aidokit/mcp-catalog listing all known MCP servers. Eight entries as of v1.0-rc2 (the seven shipped in v0.5 + graphify per ADR-0015). See reference/mcp-catalog-reference.md.
Migration script #
Code shipped under @aidokit/core/migrations that converts a project from one major spec/SDK version to another. aidokit migrate v3-to-v4 is the canonical example.
ProjectContext #
The immutable record passed to almost every adapter and stack-pack method. Built once per aidokit invocation; read-only thereafter. Defined in .docs/ARCHITECTURE.md §7.
Prereq check #
The detection-only @aidokit/prereq-check package. Reports pass / outdated / missing per prereq with per-OS install hints. Never installs anything (ADR-0008).
Role #
A bounded execution unit in the workflow: researcher, architect, planner, builder, tester-reviewer, maintainer, frontend-browser-tester (the seventh, added at Standard+).
securitySensitive #
A boolean flag on an MCP catalog entry that forces explicit confirmation on add, blocks auto-suggestion, and is unbypassable by --yes. See .docs/docs/specs/mcp-catalog.md §10.
SkillTemplate #
The data shape a stack pack returns from suggestSkills() (and what @aidokit/base-skills exports for the 18 core base skills). Carries id, name, source, content, autoLoadTriggers, preloadedBy, required, and optional description / allowedTools per ADR-0013.
Stack pack #
A package implementing the StackPack contract. Contributes detection + skills + MCP suggestions + validation commands for one tech stack. MUST NOT emit files directly.
Staging directory #
<projectRoot>/.aido-staging/. The CLI writes the full file plan here, then atomically moves into place. Deleted on success; preserved on failure. See .docs/docs/specs/security-model.md §10.2.
Trigger #
A predicate string interpreted by the MCP catalog's evaluator (e.g. always, never, stack.hasFrontend, detect.githubRemote). OR-only across the array; no AND or negation. See .docs/docs/specs/mcp-catalog.md §7.
Verb #
A user-triggered workflow command: intake, implement-task, orchestrator-next. Adapters emit them in the target CLI's native mechanism (Claude Code slash commands, Codex prompt files, Copilot prompts).
Watchdog #
The hook scripts / validators emitted by Adapter.emitWatchdog. Enforce stop conditions (test-loop, command-loop, file-edit counter, repair cap, out-of-scope writes, version bumps, tracked deletions). See .docs/docs/specs/adapter-contract.md §10.
Related Pages #
- reference/terminology.md — same set, organised by concern
- concepts/ — explanations of the key ideas