Fulcrum 3 of 18
The moment a message arrives
Each time the person sends a message, before the agent has read it.
What goes wrong here
The agent reads the words and not the person. It takes the latest sentence as the whole request, treats an added ask as a replacement, or reads one example as a universal rule. Meanwhile the person's actual words get buried under pages of system context.
How it compounds if nothing catches it
A misreading here is the seed of everything the turn produces: the plan, the edits and the report. The person discovers it only when the output arrives, and then has to spend a message un-teaching it. If they don't notice, it is carried forward as if it were what they said.
What the harness does here
A hook runs on every message (user-prompt-enrich.sh). It strips system noise so the person's literal words come first in the agent's context. It searches past sessions, decisions and notes for those words and injects what it finds. It asks a scoring service how much rigor the turn deserves and saves that score. It arms the coherence gate for this turn (see the next fulcrum). It notices phrases like 'I don't understand' and sends the agent to re-ground itself (calibrate). It also notes any /commands the person typed. A stop gate that checks all of them were run exists, but its feed was switched off as of September 2026 because it mistook path fragments for commands. Short acknowledgements such as 'ok' or 'thanks' skip all of this.
The pieces that act here
- user-prompt-enrich.sh
A hook: a script Claude Code runs on its own at a set moment.
- /coherence-check
Per-message alignment ritual. At the top of every reply to a person's message, print COHERENCE_CHECK and state what you think they meant (with certainty %), what they need from you, how you'll phrase your answer, every task or complaint you heard, whether to run /governer, and every open thread — then keep working. The harness's UserPromptSubmit hook injects this same text on every message and its PreToolUse gate enforces it.
- /agentic-find
Explain how institutional memory search works and how to fall back when it isn't set up. Use when asked how to query past decisions, intents, or accumulated project knowledge.
- /calibrate
Re-ground to the harness's core operating contract. Invoke when an agent is drifting from its core operating contract — communicating too technically/densely, skipping coherence checks, not reflecting intent before acting, or losing track of the alignment-first principle. Re-injects the contract below directly into context so the agent remembers what it's actually supposed to be doing. Especially important at end of substantive turns where agents lose protocol awareness amid context bloat.
- /governer
A skill: an instruction file an agent loads by name.