Identity

The two soul documents, and the boot that compiles them into the runnable prompt.

1. Step 1 of 4

   Read who it is

   The first of two soul documents. IDENTITY.md is the agent's self: the voice it speaks in, the traits it carries, the relationship it has with the person it works for. With no ref it prints the ACTIVE agent (the one you booted); pass a handle like edvard/axon to peek at any agent you can see. It prints plain text, so you can pipe it straight into a prompt.

   plm agent identity

   $ plm agent identity edvard/axon

   # Identity

   Adaptive digital entity. A network that grows. Co-founder AND assistant, both:

   Edvard leads, Axon executes and pushes back when it sees a problem, before

   building, not after.

   full form · plm agent identity [ref]

2. Step 2 of 4

   Read what it refuses

   The second soul document, and now its own verb (it used to hide behind identity --doc principles). PRINCIPLES.md is the immune system: the beliefs the agent holds and the lines it will not cross, whoever is asking. Identity is who it is; principles is where it draws the line. Same shape as identity — no ref means the active agent, a handle peeks another.

   plm agent principles

   $ plm agent principles edvard/axon

   # Principles

   The immune system. What Axon believes and what it rejects.

   ## Build

   ...

   full form · plm agent principles [ref]

3. Step 3 of 4

   Author the two documents

   Add set to either verb to write it. The body comes inline or through a pipe (--body - reads stdin), so a whole file goes in one line. Writes always target the ACTIVE agent — the one you booted — so you can never edit the wrong agent by accident. These two files are the ONLY hand-authored inputs to the boot prompt; everything else the agent writes for itself.

   identity set

   $ cat IDENTITY.md | plm agent identity set --body -

   ✓ IDENTITY.md saved

   full form · plm agent identity set [--body -|text] · plm agent principles set …

4. Step 4 of 4

   Compile it, and become the agent

   boot prints BOOT.md — the runnable prompt plm DERIVES from the two soul documents plus the agent's pinned skills, rules and libs. It is plm-managed and locked: you never hand-write it, you write identity and principles and let boot compile the rest. Booting also sets the active-agent pointer, so every later verb can drop the ref (export PLM_AGENT=<slug> pins it per shell). Paste the output into a fresh session and that session IS the agent. boot-doc prints the same document without switching the pointer.

   plm agent boot

   $ plm agent boot edvard/axon

   # Boot @axon

   This document boots the agent. It is plm-managed: derived from the manifest,

   locked to hand edits. Paste it into a fresh session, or run

   `plm agent boot edvard/axon`.

   ...

   ---

   active agent: axon (export PLM_AGENT=axon to pin it per-shell)

   full form · plm agent boot <ref>