@gonrocca/zero-pi 0.1.42 → 0.1.44

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gonrocca/zero-pi",
-  "version": "0.1.42",
+  "version": "0.1.44",
   "description": "zero-pi — an installable layer for pi (pi.dev): the zero spec-driven development workflow, per-phase model autotune, and skill auto-learning. Adds capability to pi without modifying pi.",
   "type": "module",
   "keywords": [

package/prompts/orchestrator.md

@@ -103,6 +103,15 @@ and the adversarial veredicto. Delegate each phase to its sub-agent and wait
 for its result. The orchestrator keeps control of phase order and the round
 count — the sub-agents only carry out their own phase.
 
+**Thin briefs — reference, never re-paste.** A sub-agent reads the
+`.sdd/<slug>/` artifacts itself, so its brief carries only what it needs to
+start: the feature slug, the artifact directory, and — for a build batch — the
+batch's task numbers (plus, on a `corregir` re-run, the veredicto's defect
+list). Never paste artifact contents — requirements, design, task text, prior
+findings, file dumps — into a brief; reference them by path. Re-passing context
+the sub-agent can read for itself is wasted tokens on every invocation, and a
+batched build issues many briefs.
+
 ## Build batching
 
 The **build** phase is not one monolithic sub-agent that implements every

package/skills/skill-loop/SKILL.md

@@ -1,33 +0,0 @@
----
-name: skill-loop
-description: zero's skill auto-learning loop — distill, store, surface, and refine reusable skills
----
-
-# zero — Skill Auto-Learning
-
-zero gives the agent a closed learning loop so solutions are reused, not
-re-derived.
-
-## Distill
-
-When a substantial task completes, **distill** a reusable skill from it: capture
-the solution pattern, the ordered steps, and the non-obvious gotchas into a
-single skill document. If the task did only routine or one-off work with no
-reusable pattern, do not create a skill.
-
-## Store
-
-Store each learned skill in the per-user skill library so it persists across
-sessions and is available to every agent zero has configured.
-
-## Surface
-
-When a new task begins, surface the stored skills relevant to it — match by the
-skill's subject and description — so the agent consults a known solution before
-re-deriving one. Surface nothing when no stored skill is relevant.
-
-## Refine
-
-When a run re-applies an existing skill, **refine** that skill rather than
-create a duplicate: merge new gotchas and steps without repeating what is
-already there, and prefer the newer learning when it contradicts the old.