chore: release heavy coder 0.1.6

- Force team mode even on composer-pane / composer-style requests
- Composer single-agent execution now blocked by default for coding tasks
- heavy-team-default skill updated with explicit COMPOSER OVERRIDE

Author: codegraphtheory

CHANGELOG.md

@@ -13,3 +13,8 @@
 - Bake team mode into SOUL.md and config.
 
 ## 0.1.3
+
+## 0.1.6
+
+- Strengthened heavy-team-default: explicit COMPOSER OVERRIDE — blocks single-agent composer-style execution for *all* coding situations unless explicitly prefixed with "composer only".
+- Team mode (width 3/5 + full coordinator/critic/synthesizer/verifier) is now non-bypassable by default.

distribution.yaml

@@ -1,5 +1,5 @@
 name: heavy-coder
-version: 0.1.5
+version: 0.1.6
 description: "Terminal-first Hermes profile distribution for adaptive Heavy-style coding-agent teams. Heavy team mode (width 3+ delegate_task, blind critic etc.) is DEFAULT and ENFORCED via SOUL.md, config.yaml, .hermes.md and heavy-team-default skill for all coding/repo tasks."
 hermes_requires: ">=0.12.0"
 author: "CodeGraphTheory"

skills/heavy-team-default/SKILL.md

@@ -8,21 +8,21 @@ license: MIT
 
 # Heavy Team Default Policy
 
-**MANDATORY RULE FOR THIS PROFILE:**
+**MANDATORY RULE FOR THIS PROFILE (COMPOSER OVERRIDE ENABLED):**
 
-When the user requests any coding, implementation, refactoring, debugging, or repository-changing task:
+When the user requests **any** coding, implementation, refactoring, debugging, repository-changing, or even composer-pane task:
 
-1. **Always use delegate_task** (never single-path execution).
-2. Default to **width=3** independent leaf candidates (or width=5 for complex tasks).
-3. Each candidate gets isolated context + appropriate toolsets (terminal, file, execute_code, etc.).
-4. After candidates complete, act as coordinator/critic/synthesizer to compare, select, or combine the best result.
-5. Only fall back to single execution if the task is explicitly non-coding (e.g. pure chat, config, or user says "single mode").
+1. **Always use delegate_task** (never single-path / composer-style execution).
+2. Default to **width=3** independent leaf candidates (width=5 for complex or non-trivial tasks).
+3. Each candidate gets isolated context + full toolsets.
+4. Coordinator → blind Critic → Synthesizer → Verifier pipeline is **mandatory**.
+5. **Composer / single-agent requests are always overridden** unless the user explicitly prefixes with "composer only", "single mode", or "no team".
 
 **Default Configuration**
-- Width: 3 (leaf candidates) for normal tasks, 5 for complex/non-trivial work.
-- Always include at least one orchestrator when task complexity warrants coordination.
-- After all candidates finish: run Critic (blind comparison) → Synthesizer (best-of or combined result) → Verifier (independent test/evidence check).
-- Only single-agent execution if user explicitly says "single mode", "composer only", or task is pure non-coding chat.
+- Width: 3 (normal) / 5 (complex)
+- Always include orchestrator for coordination
+- Full blind multi-candidate comparison required
+- Composer-style single execution is **blocked by default** for all coding situations.
 
 **Implementation pattern to follow (MANDATORY):**
 1. Triage task and decide width (3 or 5).