instar 1.3.473 → 1.3.474

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "instar",
-  "version": "1.3.473",
+  "version": "1.3.474",
   "description": "Coherence infrastructure for self-evolving AI agents — on the Claude Code or Codex subscription you already have.",
   "type": "module",
   "main": "dist/index.js",

package/src/data/builtin-manifest.json

@@ -1,8 +1,8 @@
 {
   "$schema": "./builtin-manifest.schema.json",
   "schemaVersion": 1,
-  "generatedAt": "2026-06-10T18:12:02.588Z",
-  "instarVersion": "1.3.473",
+  "generatedAt": "2026-06-10T18:39:03.415Z",
+  "instarVersion": "1.3.474",
   "entryCount": 201,
   "entries": {
     "hook:session-start": {

package/upgrades/1.3.474.md

@@ -0,0 +1,32 @@
+# Upgrade Guide — vNEXT
+
+<!-- assembled-by: assemble-next-md -->
+<!-- bump: patch -->
+
+## What Changed
+
+Added `docs/AGENT-AUTONOMY-PRINCIPLES.md` — a verbatim capture of the operator's two
+foundational principles for how Instar agents should operate: (1) almost all "blockers" are
+false blockers — judgment calls to work through an authority → access → dry-run → codify
+pipeline rather than walls; (2) the spec-design process should frontload every user decision
+so the agent completes the spec in a SINGLE autonomous run, and mid-run decisions are
+cheap-to-change-after rather than stop-and-wait. No runtime code changed; this is the source
+document for an in-progress exploration of integrating these into Instar fundamentals
+(hooks, skills, and possibly new Constitutional standards).
+
+## What to Tell Your User
+
+Nothing changes in how the agent runs today. This records, word-for-word, two operating
+principles the operator laid out — so future work (and any new constitutional standards
+derived from them) traces back to a single durable source instead of living only in a chat
+log.
+
+## Summary of New Capabilities
+
+- A new reference doc: the verbatim, citable source of the false-blocker pipeline and the
+  decision-frontloading principle.
+
+## Evidence
+
+- New doc only; no source change. The two principles are quoted verbatim with the operator's
+  attribution, date, and topic, followed by a clearly-labeled derivation (the quote governs).

package/upgrades/side-effects/agent-autonomy-principles.md

@@ -0,0 +1,36 @@
+# Side-Effects Review — Agent Autonomy Principles doc
+
+**Version / slug:** `agent-autonomy-principles`
+**Date:** `2026-06-10`
+**Author:** `echo`
+**Tier:** `1` (new docs file only; no runtime code, no decision logic)
+**Second-pass reviewer:** `not required`
+
+## Summary of the change
+
+Adds `docs/AGENT-AUTONOMY-PRINCIPLES.md`, a verbatim capture of the operator's two
+foundational autonomy principles, with a clearly-labeled derivation section (the quote
+governs). Pure documentation — no `src/`, no behavior, no decision logic.
+
+## Decision-point inventory
+
+None. The change adds no code and no decision logic — it is a reference document.
+
+## 1. Over-block
+
+Nothing is rejected at runtime. The doc changes no gate, message path, or API.
+
+## 2. Under-block
+
+The doc does not, by itself, enforce the principles — integration into hooks/skills/standards
+is the explicit follow-up (the exploration topic). It is the source, not the mechanism.
+
+## 3. Level-of-abstraction fit
+
+Right layer: a standalone `docs/` reference that the future constitutional/standards
+derivations cite, rather than pre-emptively editing `docs/STANDARDS-REGISTRY.md` before the
+exploration has decided the shape.
+
+## Migration / rollback
+
+No migration (docs only). Rollback = delete the file.