equip-agentslisted

# Equip an agent() with skills, tools, MCP, and memory The engine's built-in coding tools (read, write, edit, ls, grep, glob, bash, webfetch, and more) are on by default, so a plain `agent(prompt)` already works in the run's workspace. Everything beyond that is added **per `agent()` call**, on its options object. None of it is declared in `workflow.jsonc`: two `agent()` calls in the same workflow can carry different skills, tools, and memory. ## The workflow package A workflow is a **package**: the directory holding the descriptor and the entry. Point `boardwalk check`/`build`/`deploy` at the directory. ```text code-review/ workflow.jsonc # the deployment descriptor package.json README.md # the workflow's landing page in the dashboard src/index.ts # the entry: exports run src/lib/parse.ts # helper modules the entry imports normally skills/ reviewer/ SKILL.md # a reusable procedure checklist.md # a bundled resource the skill can pull on demand ``` You never list source files: the code that ships is whatever the entry imports. `skills/**` and `README.md` ride along by convention; other non-code assets (prompt templates, fixtures) ship via the `files` allowlist in `workflow.jsonc`. The `skills/` folder is not imported; it is bundled and resolved by name at run time (below). ## Give an agent reusable skills `agent(prompt, { skills: ["reviewer"] })` makes `skills/reviewer/SKILL.md` available to that