Lecture 07. Silent harness faults and configuration fitness checks

Part 3 · The Harness as a Control Layer · Scheduled 20 Dec 2026 · 1-hour lecture

Upcoming course outline. This class has not been recorded as studied.

Learning objectives

Agent = Model + Harness; the six harness components (Day 1 pp. 26-28) mapped onto the student's own part2-ref: wrapper and cap = gateway and budget, step cap = stop condition, MCP = tools, context policy = context, scripted-LM tests = sensors.
Guides vs sensors; settings precedence; hooks and tool guards fail open unless set to block; CI configuration checks as architecture fitness functions.
Static vs dynamic context; skills and progressive disclosure; four failure modes; the 90% trigger bar (Day 3 p. 22).
'Isolation is a trap' (Day 3 p. 25); the ratchet (Osmani).

Industry: Teams lose days to a hook that never fires; students learn to prove what loaded instead of trusting the config.
Academic: Skill-trigger accuracy with distractors is a measurable property; the method transfers to any prompt-routing study.
Elsewhere: Students diagnose their own coding-agent setup at home and at work.

These are preparation requirements in the draft; staff materials are not bundled with this site.

The linked reference library contains the source URLs extracted from the course PDF. Availability notes above reflect the draft of 4 November 2026.

Continue with Lab 07: Silent harness faults and configuration fitness checks. Every lab contributes an artifact to the same repository and a measurement to the results ledger.

[Lecture notes and observations]