{ From Chat to Coworker }
A nine-notebook curriculum that walks an LLM from a chat window to a federally-auditable coworker — prompts, the loop, skills, identity, audit, runtime defenses, spec-driven workflow, and a live capstone.
the arc:
- 01 chat_to_coworker — raw LLM calls. Model is interchangeable; the harness is what matters.
- 02 prompts_that_work — five anti-patterns and the plan · do · check loop that makes agents self-correcting.
- 03 the_loop — one tool, streaming events. The loop is just LLM + tools + memory.
- 04 skills_and_prompts — SKILL.md folders. The agent loads only what it needs.
- 05 identity_aligned_decisions — same data, different identity, different decision. Tribal knowledge under git.
- 06 full_agent_audit — federal-grade tamper-evident chain. Which version of which identity made which decision.
- 07 security_at_runtime — PII redaction, sandboxed exec, allowlists, budgets. OWASP LLM/ASI top-10 mapped.
- 08 coding_workflow — spec-driven development. Plan/Do/Validate is fractal — same pattern at three scales.
- 09 your_workflow — capstone. Pick a real lab task, build the harness around it live.
stay in transmission range:
Drop your email and we'll send you the access key, the slide decks when they ship, and post-workshop write-ups. No marketing spray — roadshow follow-ups only.
have an access key already?
Type roadshow --key loop-… in the
terminal below, or open any of the materials pages and enter the key
in the gate form. Your unlock persists in this browser.