Glossary
High-frequency terms used across this guide.
Agent. A skill with a persona and a menu. Eight ship with Foundry — see the agent quick-reference. There is no separate agent file format: invoke the skill, the skill becomes the agent.
Alloy. The @engineering11/alloy npm package. A versioned knowledge corpus of SDK docs, patterns, conventions, infrastructure guides, and the SDK dependency matrix. Foundry skills consult it through skill:e11-knowledge before any platform-touching decision.
Artifact. Anything a skill produces — a brief, a PRD, an architecture spine, a story file, a spec, a test suite. Artifacts land in {planning_artifacts}/, {implementation_artifacts}/, or {project_knowledge}/ depending on type.
Catalog. .foundry/_config/e11-help.csv — the assembled manifest of every installed skill, one row per command. E11 Master renders its menu from it; e11-help reasons over it. Source: src/foundry/module-help.csv.
Central config. Identity and project-level settings (your name, languages, artifact paths, the agent roster), resolved through four layers by .foundry/scripts/resolve-config.mjs: config.toml / config.user.toml (installer-owned) then custom/config.toml / custom/config.user.toml (yours). See Customization.
Ceremony dial. Foundry's sizing principle: oneshot e11-dev-flow for tiny changes, the quick engine for small work, the PRD fast path for features, the full lifecycle for products, e11-dev-auto for unattended runs. See Quick flow.
Citation. A path-style reference Foundry skills attach to non-obvious claims so you can trace where the claim came from. Two schemes: alloy://<path> (claim grounded in alloy) and sdks-nodejs-skill://<file> (claim grounded in an upstream-skill file).
Command code. The 2–3 letter code shown in menus ([PR] PRD, [DS] Dev Story, …). Stable across installs, unique across the catalog.
customize.toml. The data surface a skill ships — persona, menu, templates, paths, gates, review layers — as TOML defaults. You override it from .foundry/custom/; you never edit the shipped file.
E11 Master. The default entry point — presents the live catalog, routes any request to the right skill or agent, and executes directly when no specialist fits.
.foundry/. Support files the installer writes: central config, your custom/ overrides, the resolver scripts, the assembled catalog, and the installed module content. Skills themselves live in your tool's skill directory (.claude/skills/, .agents/skills/, .opencode/skills/).
.foundry-output/. The default artifact root (configurable via the output_folder prompt). Holds planning-artifacts/ and implementation-artifacts/.
Implementation artifacts. Output from Phase 4 and the quick flow: sprint status, stories, specs, code reviews, retrospectives. Default location: {output_folder}/implementation-artifacts/.
Module. A bundle of skills the installer can include or exclude. Foundry is itself a module; others can be installed alongside it.
Override layers. The three per-skill files resolved at activation: <skill>/customize.toml (defaults) ← .foundry/custom/<skill>.toml (team, committed) ← .foundry/custom/<skill>.user.toml (personal, gitignored). Sparse overrides only — see Customization.
Phase. One of the five delivery stages (Analysis → Planning → Solutioning → Implementation, plus the off-cycle Scaffold) lifecycle skills are grouped under. Anytime tools sit outside the phases.
Planning artifacts. Output from Phases 1–3: briefs, PRDs, UX designs, architecture spines, epics. Default location: {output_folder}/planning-artifacts/.
Principles. Action-oriented rules an agent always follows, shipped in its customize.toml (e.g. Dev: "All existing and new tests pass 100% before a story is ready for review."). Your overrides append to them.
Project knowledge. Long-lived reference material — research output, document-project results, project context. Default location: docs/ (configurable via the project_knowledge prompt). Distinct from artifacts because it's reference, not deliverable.
Query key. The colon-separated string passed to skill:e11-knowledge to fetch an alloy doc. Examples: sdk:auth, api:datastores, pattern:multi-tenancy, matrix:unit, list:sdks. See 04-knowledge-layer.md.
Quick flow. The lightweight path for small single-contributor work — e11-dev-flow (with e11-spec-flow when you want the contract first). Bypasses PRD/architecture/story ceremony. See Quick flow.
Roster. The agent cast as data: one [agents.<code>] table per agent in central config, written from module.yaml at install time. Roster-aware skills (e11-multi-perspective, e11-retrospective, e11-refine) resolve their cast from it, and you can add personas via .foundry/custom/config.toml.
Skill. The unit of everything in Foundry: a directory with a SKILL.md, a customize.toml, and optional step files and templates. Agents, workflows, and anytime tools are all skills.
Spec. The contract the quick flow works from — a SPEC.md kernel (plus companions) produced by e11-spec-flow or derived inline by e11-dev-flow. Its frontmatter status is the state machine: draft → ready-for-dev → in-progress → in-review → done.
Sprint board. {implementation_artifacts}/sprint-status.yaml — the development_status map the story cycle reads and syncs. Owned by e11-sprint-planning; orchestrators read it, the sprint/story skills write it.
Step file. A .md file under a workflow's steps/ directory (e.g. step-02-review.md). Multi-stage workflows load exactly one at a time, just-in-time.
Story. The implementation unit. Prepared by e11-create-story ([CS], [VS] to validate), executed by e11-dev-story ([DS]), reviewed via e11-code-review ([CR]). Lives under {implementation_artifacts}/.
Workflow. A lifecycle skill tied to a phase folder. Multi-stage workflows carry a steps/ directory and are deterministic enough to pause and resume.