Series · roadmap stage 13/13
The Agentic Series
Ten lessons in order, from the mental model to a capstone checklist. Every lesson is structured by us and grounded in a real, cited source from the registry — no invented commands.
10 of 10 lessons published
- Lesson 1 What makes an agent — model, tools, loop, guardrails
Build the map before touching anything: an agent is a model that reasons, tools that act, a loop that iterates, and guardrails that bound it. Meet a real data-engineering agent from the registry, point at each of the four parts, and see whe
- Lesson 2 Onboard an agent with the DE Data Toolkit
Onboard an agent onto a data repo with the DE Data Toolkit plugin: the SQL/Spark review skills, the destructive-SQL guard, a secret scanner, and the file/git MCP servers already wired. Learn the three commands and which hooks run underneath
- Lesson 3 Inside a skill — SKILL.md, references, examples, the loading protocol
A skill is the unit of reusable agent behavior: one SKILL.md with front-matter, plus directories the agent loads on demand. Dissect a real registry skill file by file, learn the evaluation rubric its objectives set, then install it and comp
- Lesson 4 Audit an MCP server's tool contracts
Before connecting an MCP server, audit its tools for blast radius. Read each tool schema, classify every tool as Read, Write, Destructive, Spend or Exfiltrate, and produce a least-privilege connection config that grants only what the task n
- Lesson 5 Hooks — deterministic gates around an agent
The model can be talked into anything; a hook is code on the tool path. Walk the four safety hooks every data repo should know — destructive-SQL guard, prod write confirmation, secret scan, backfill safety check — and wire the two the start
- Lesson 6 Keep an agent read-only until a human approves
Coding agents can do real data-engineering work, but can also ship a schema change or drop a table. Walk the Agent-in-the-Loop Review pattern: separate read/recommend from write/execute, put a human gate before anything irreversible, and en
- Lesson 7 Wire the core MCP servers for a data repo
Wire the three reference MCP servers a data repo needs: filesystem for the working directory, git for history and commits, and GitHub for issues and pull requests. Each connection is scoped — one directory, one repository, one fine-grained
- Lesson 8 Production gotchas — untrusted content, destructive actions, permission scope
An agent executes with your credentials, so its context and config are an attack surface. Read the MCP security model, treat fetched content as untrusted input, keep destructive statements deny-by-default, and scope every credential to the
- Lesson 9 When an agent does the wrong thing — contain, audit, postmortem
An agent merged a change it should not have. Roll back first, use the scoping you configured to contain and reconstruct what happened — diffs, logs and approve-once records — walk the incident workflow, and close with a blameless postmortem
- Lesson 10 Capstone — the guarded agent checklist
Run the whole series against one agent setup: skills that load on demand, tools scoped at the connection, least-privilege credentials, audited tool contracts, hooks that fail closed, and a human gate on anything irreversible. Prove each box