is-shapelisted

# Shape the `_agent/` Canonical protocols: read `${CLAUDE_PLUGIN_ROOT}/reference/form-primitive.md` for procedures, checklists, routines, and agents; read `${CLAUDE_PLUGIN_ROOT}/reference/form-perspective.md` for reusable thinking patterns. ## Which one - **Primitive** — *how to work* in a situation: a procedure, checklist, review pattern, memory routine. Any part of `_agent/` with `name` + `description` frontmatter (the description is the trigger). Use the canonical form-primitive reference above. - **Perspective** — *how to think* about something repeatably: an evaluation or analysis with Object Definition, Thinking Structure, Expected Outcome. Use the canonical form-perspective reference above. If the user wants consistent *evaluation/analysis*, it's a perspective. If they want a repeatable *procedure/behavior*, it's a primitive. ## How Elicit progressively — don't demand the full structure up front. Start from a real instance ("walk me through the last time you did this"), find the invariant, draft, and **show it before saving**. Primitives and perspectives live in `_agent/` at the level where they apply; everything in `_agent/` composes along the path root → current position. Use native `Write` / `Edit` for `_agent/skills/*`, `_agent/perspectives/*`, and other agent-context primitives because they use `name` + `description` frontmatter, not Note `name` + `summary` frontmatter. For a skill, its entry id and frontmatter `name` must be the same portable id — follow