@cubis/foundry 0.3.72 → 0.3.73

package/CHANGELOG.md

@@ -4,6 +4,14 @@ All notable changes to this project are documented in this file.
 
 ## [Unreleased]
 
+## [0.3.73] - 2026-03-13
+
+### Fixed
+
+- Fixed skill loading protocol across all platforms (Codex, Copilot, Claude, Antigravity): agents now load skills recommended by `route_resolve` (`primarySkillHint`, `primarySkills[0]`, `detectedLanguageSkill`) before executing non-trivial tasks instead of ignoring them.
+- Added TIER 1b to the managed skill-routing block so `cbx workflows sync-rules` propagates the post-route skill loading rule to installed projects.
+- Corrected the Codex sandbox override note to explicitly name the Foundry MCP tools and the expected post-route loading behavior.
+
 ## [0.3.72] - 2026-03-13
 
 ### Fixed
@@ -287,5 +295,3 @@ All notable changes to this project are documented in this file.
 - Added centralized Postman config via `cbx_config.json`.
 - Added MCP scope controls (`project/workspace/global/user`) with platform-aware runtime placement.
 - Kept rules/engineering artifacts in workspace scope while defaulting skills/powers install to global scope.
-
-

package/dist/cli/core.js

@@ -3193,10 +3193,11 @@ function buildManagedWorkflowBlock(platformId, workflows) {
     lines.push("");
     lines.push("<!-- cbx:managed:skill-routing start -->");
     lines.push("Classify intent before any MCP call.");
-    lines.push("- TIER 1 DIRECT: `skill_get <exact-skill-id>` when domain and skill are obvious.");
-    lines.push("- TIER 2 TARGETED SEARCH: one `skill_search <1-3 word noun>` max, then `skill_validate` -> `skill_get`.");
-    lines.push("- TIER 3 SKIP: no MCP call for conversational, same-session, or native-tool-sufficient work.");
-    lines.push("- Never pre-load skills or agents speculatively.");
+    lines.push("- TIER 1 DIRECT: `skill_get <exact-skill-id>` when skill ID is known from route or context.");
+    lines.push("- TIER 1b ROUTE-RECOMMENDED: after `route_resolve`, load `primarySkillHint` or first `primarySkills` entry via `skill_validate` -> `skill_get` before executing non-trivial tasks.");
+    lines.push("- TIER 2 TARGETED SEARCH: one `skill_search <1-3 word noun>` max when domain is unclear, then `skill_validate` -> `skill_get`.");
+    lines.push("- TIER 3 SKIP: no MCP call for conversational replies, trivial one-liners, or identical skill already loaded this session.");
+    lines.push("- Never pre-load skills or agents speculatively before route resolution.");
     lines.push("- Keep one primary agent and one primary skill by default.");
     lines.push("- Add supporting skills only when the active task explicitly crosses domains.");
     lines.push("- Direct skill-package creation or repair work to `skill-creator` instead of starting with `skill_search`.");