Fulcrum 1 of 18
Before the harness knows the person
Installation and the first sessions of someone new, before the harness has any record of what this person wants.
What goes wrong here
Much of what the harness does for Jonathan draws on a record of what he wants: past sessions, confirmed decisions, principles mined from his corrections, and notebooks that predict his reactions. A new person has none of that. If the harness is installed as-is, the pieces that consult that record come back empty, or they answer from someone else's history. The agent falls back to guessing about the person, which is exactly what the harness is meant to reduce.
How it compounds if nothing catches it
Every later fulcrum that leans on memory is weaker from day one. That includes looking up context before acting, predicting what the person would say instead of interrupting them, and recognising what was already decided. It is worse if a piece quietly carries Jonathan's own examples or notebooks, because then the new person's agent aligns to the wrong human.
What the harness does here
This part is not built yet. It is the planned first-run experience: one command to install and a setup that says what it will do before doing it, so building a synthetic understanding of the new person's intent is part of the workflow from the start rather than assumed. Each checkpoint offers options: for example, walking the person through creating their own NotebookLM notebooks (the 'oracles'), or switching that piece off. There are on/off switches, and intensity dials where something could fire too often. Part of that workflow has to build a first map of the new person's intent. Some existing pieces suggest how. align works out and records confirmed intent. flywheel-consultant describes a method for building oracles from one's own sessions. predict-required-skills2 and instinct-harvest have capture loops that can start from nothing.
The pieces that act here
- /align
Figure out what the human actually wants before doing anything. Use when starting a session, when intent is unclear, when switching topics, or when the human says "/align". This is not a technical document — it's a guide for how to listen.
- /intent-architect
Curate and map human intent into validated seed documents with nested structure, certainty scores, and predict-and-validate communication
- /seed
Load the current confirmed intent seed for this project into session memory, so alignment work starts from the person's stated intent rather than a guess. If none exists yet, say so plainly and point to /align, which is what creates one.
- /flywheel-consultant
Expert on self-improving agent flywheels — systems where session data gets extracted into structured oracles that steer future sessions better. An alignment flywheel built this way is the first working specimen of this pattern. Use this skill to understand how it works, audit signal quality, extend it to new data sources, diagnose degradation, OR to abstract the pattern and build a flywheel for a completely different domain. This is a case study AND a template.
- /predict-required-skills2
Predicts the skill sequence a new scope of work needs, grounded in why past sessions actually reached for each skill — not a guess. Works two ways: a local fallback that reads your own past session transcripts (no setup), or an optional NotebookLM oracle built from an automatically-growing corpus of your own \"why I used this skill\" annotations. Invoke at the start of any distinct scope to get the skill sequence, in order, grounded in cited precedent.
- /instinct-harvest
Mine INSTINCTS (not facts) from a person's confirmed sanity checks and coherence checks in session transcripts, and maintain a per-domain instinct ledger that agents read before entering a situation. General tier — domain-agnostic, works standalone with a built-in "general" domain, or a specialized domain layer can supply its own ledger path and situation taxonomy. Use when seeding a domain's instinct ledger, when harvesting a session's confirmed checks, when generating the ledger index, or when checking whether the loop is healthy.