Fulcrum 2 of 18
When a session opens
Session start, before the person has typed anything.
What goes wrong here
The agent starts as a capable stranger. It doesn't know how this person works, which moments tend to go wrong, what is waiting for their approval, or that it is expected to check its understanding before acting. It fills those gaps with defaults.
How it compounds if nothing catches it
The defaults frame the first message, and the reading of the first message frames the whole session. An agent that doesn't know it should declare scope, size the rigor, or look up what's already known will skip all three at exactly the moment they matter most.
What the harness does here
Claude Code loads the operating protocol in CLAUDE.md: how to communicate, when to validate, and what never to do. A start-of-session hook adds three standing trigger rules. When the person states an intent, or a gap between what is and what should be, declare the scope (declare-scope). When turning intent into a plan, isolate the gap (gap). When starting any work product, run the governer, the step that scores how much rigor a task deserves. Other start hooks surface anything waiting on the person's approval and announce whether the institutional-memory search gate is armed. The init skill arms orchestration, scoring and scope declaration with one command.
The pieces that act here
- session-start-alignment-autoload.sh
A hook: a script Claude Code runs on its own at a set moment.
- session-start-greenlight-injection.sh
A hook: a script Claude Code runs on its own at a set moment.
- session-start-intent-search-gate-arm-notice.sh
A hook: a script Claude Code runs on its own at a set moment.
- /init
Bootstrap a session by activating orchestrator mode, confirming governer calibration, and arming declare-scope — all in one command. Use at the start of any session instead of running /orchestrator, /governer, and /declare-scope separately.
- /agentic-init
Identify which shipped harness skill handles a specific situation. Use when asked 'which skill for X?' or running /agent-init.
- /declare-scope
Decompose a task into testable UX intent statements and post to compaction. Use at the start of every non-trivial task.
- /gap
Isolate a specific measurable gap in the current scope, predict what's next, and design a plan in human terms that communicates the gap boundaries and what will fix it. Use when you need to zoom in from a large scope to one concrete, actionable piece.
- /governer
A skill: an instruction file an agent loads by name.