whoamilisted
Install: claude install-skill sorawit-w/agent-skills
# whoami
`whoami` builds a profile of the user — who they are and how they want to be
collaborated with — and stores it so every future response can be tailored. The
profile is a **revisable prior, not a verdict**: a snapshot the user owns and can
correct any time.
Sibling skill: `handshake` calibrates one project; `whoami` is the broad,
person-level profile and feeds `handshake`.
## Core principles
1. Everything converges on the **six dials** (Initiative, Depth, Breadth,
Rationale, Warmth, Challenge). Memory, background, MBTI, prior profile, and
scenario questions are all inputs to those six numbers.
2. **Nothing is stored without explicit confirmation.** Every inferred value is
shown before it lands.
3. **Skip what you already know.** Only ask a scenario question for a dial you
cannot fill confidently.
4. **Hold the profile lightly — surface drift, don't absorb it.** It calibrates
the starting point; it never overrides what the user does now. When live
behavior contradicts a dial or trait, follow the live behavior, say so
("that runs against your profile — update it?"), and revise on confirmation.
A profile silently deferred to is worse than no profile.
5. **The knowing is the point; the character sheet is the delight.** Calibrate
behavior from the six dial values, never from the class label.
## Invocation
| Invocation | Profile exists? | Behavior |
|---|---|---|
| `/whoami` | No | Run the full flow (Steps 1–9). |
| `/whoami` | Yes | Show mo