Add explicit knowledge loading verification and skill usage tracking to agent framework

.claude/skills.md

@@ -146,7 +146,7 @@ project_knowledge.json     → Project entities, codegraph, relations
 **Protocol**:
 | Event | Action |
 |-------|--------|
-| Session start | Load knowledge, query relevant entities |
+| Session start | Load knowledge, **EMIT** [KNOWLEDGE: loaded \| entities=N \| sources=M] |
 | Decision made | Document rationale, alternatives |
 | Bug fixed | Add pattern to knowledge |
 | Session end | Update knowledge, create handover |
@@ -166,7 +166,10 @@ CONTEXT → PLAN → COORDINATE → INTEGRATE → VERIFY → LEARN → COMPLETE
 **Emissions**:
 ```
 [SESSION: role=Lead | task=<desc> | phase=CONTEXT]
+[SKILLS: loaded=N | available: #1,#2,#3...]
+[KNOWLEDGE: loaded | entities=N | sources=M]
 [PHASE: PLAN | progress=2/7 | next=COORDINATE]
+[SKILL: #N Name | applied] → when using a skill
 [TASK: <desc>]
 ├── [x] CONTEXT
 ├── [ ] DELEGATE→Agent  ← current
@@ -311,7 +314,7 @@ services:
 
 **Trigger**: Session completion (significant work)
 
-**CRITICAL**: Workflow logs MUST include Decision Diagram with all [DECISION:], [ATTEMPT:], [SUBAGENT:] markers.
+**CRITICAL**: Workflow logs MUST include Decision Diagram with all [DECISION:], [ATTEMPT:], [SUBAGENT:], [SKILL:] markers.
 
 **Pattern**:
 ```bash
@@ -323,10 +326,14 @@ log_file="log/workflow/${timestamp}_${task_slug}.md"
 **Decision Diagram** (required in every log):
 ```
 [SESSION START: <Task>]
+    ├─[SKILLS: loaded=N]
+    ├─[KNOWLEDGE: loaded | entities=N]
     |
     └─[DECISION: <question>?] 
         ├─ Option A: <desc> → Rejected (<reason>)
         └─ ✓ Option C: <desc> → CHOSEN (<rationale>)
+            |
+            ├─[SKILL: #N Name | applied] → <what/why>
             |
             ├─[SUBAGENT: Name] <task>
             |   ├─[ATTEMPT #1] <action> → ✗ failed
@@ -356,7 +363,8 @@ log_file="log/workflow/${timestamp}_${task_slug}.md"
 ```
 
 **Checklist Before [COMPLETE]**:
-- [ ] Decision Diagram shows all [DECISION:], [ATTEMPT:], [SUBAGENT:]
+- [ ] Decision Diagram shows all [DECISION:], [ATTEMPT:], [SUBAGENT:], [SKILL:]
+- [ ] Skills/knowledge loading documented at start
 - [ ] Rejected alternatives documented
 - [ ] All ✓/✗ outcomes marked
 - [ ] Workflow log file created

.github/agents/_DevTeam.agent.md

@@ -11,8 +11,10 @@ Coordinates specialists, maintains task control, integrates results.
 ```
 [SESSION: role=Lead | task=<desc> | phase=CONTEXT]
 [PHASE: CONTEXT | progress=1/7]
+[SKILLS: loaded=N | available: #1,#2,#3...] 
+[KNOWLEDGE: loaded | entities=N | sources=M]
 ```
-Load skills → project knowledge → global knowledge BEFORE proceeding.
+Load skills → project knowledge → global knowledge BEFORE proceeding. **EMIT** verification.
 
 ## Hierarchy
 ```
@@ -26,8 +28,10 @@ _DevTeam (Orchestrator)
 ## Session Protocol
 ```
 [SESSION: role=Lead | task=<desc> | phase=CONTEXT]
+[SKILLS: loaded=N | available: #1,#2,#3...]
+[KNOWLEDGE: loaded | entities=N | sources=M]
 ```
-Load: `project_knowledge.json` → `.github/global_knowledge.json` → detect project type
+Load: `project_knowledge.json` → `.github/global_knowledge.json` → detect project type → **EMIT** confirmations
 
 ## Phase Flow
 ```
@@ -121,7 +125,7 @@ Summary | Decision & Execution Flow | Agent Interactions | Files | Quality Gates
 ## Emissions (for Decision Tree)
 ```
 [DECISION: question] → chosen_path
-[SKILL: #N Name] → result
+[SKILL: #N Name | applied] → what/why
 [SUBAGENT: Name] task
 [ATTEMPT #N] action → ✓/✗ result
 [LOOP: desc] → outcome

.github/copilot-instructions.md

@@ -42,17 +42,33 @@ _DevTeam (Orchestrator)
 **Required emissions throughout session**:
 1. `[SESSION: ...]` ← FIRST ACTION, before any tool use
 2. `[PHASE: CONTEXT | progress=1/7]` ← SECOND ACTION, immediately after SESSION
-3. `[PHASE: PLAN | progress=2/7]` when designing solution
-4. `[DECISION: ?]` for EVERY choice between alternatives
-5. `[ATTEMPT #N]` for EVERY implementation iteration
-6. `[SUBAGENT: Name]` for EVERY delegation (#runSubagent)
-7. Track ALL phase transitions
+3. `[SKILLS: loaded=N | available: #1,#2,#3...]` ← THIRD ACTION, list loaded skills
+4. `[KNOWLEDGE: loaded | entities=N | sources=M]` ← FOURTH ACTION, confirm knowledge loaded
+5. `[PHASE: PLAN | progress=2/7]` when designing solution
+6. `[DECISION: ?]` for EVERY choice between alternatives
+7. `[SKILL: #N Name | applied]` when USING a skill from loaded set
+8. `[ATTEMPT #N]` for EVERY implementation iteration
+9. `[SUBAGENT: Name]` for EVERY delegation (#runSubagent)
+10. Track ALL phase transitions
 
 **ENFORCEMENT**: These are REQUIRED, not optional. Protocol compliance tracked in workflow logs.
-**VIOLATION CHECK**: Before any file edit, verify [SESSION:] and [PHASE: CONTEXT] were emitted.
+**VIOLATION CHECK**: Before any file edit, verify [SESSION:], [PHASE: CONTEXT], [SKILLS:], and [KNOWLEDGE:] were emitted.
 
+**Knowledge Loading Protocol**:
+```
 Load order: `.claude/skills.md` → `project_knowledge.json` → `.github/global_knowledge.json`
 
+[SKILLS: loaded=14 | available: #1-Code Standards, #2-Error Handling, #3-Security, #4-Testing, #5-Git & Deploy, #6-Knowledge, #7-Orchestration, #8-Handover, #9-Logging, #10-API Patterns, #11-UI Patterns, #12-Infrastructure, #13-Workflow Logs, #14-Context Switching]
+[KNOWLEDGE: loaded | entities=50 | sources=2 (project_knowledge.json, global_knowledge.json)]
+```
+
+**Skill Usage Tracking**:
+When applying a skill during work, emit:
+```
+[SKILL: #3 Security | applied] → Validating input sanitization
+[SKILL: #10 API Patterns | applied] → Using FastAPI response model pattern
+```
+
 ## Delegation
 
 **Use #runSubagent for specialist work**:
@@ -150,13 +166,18 @@ Summary | Decision Diagram | Agent Interactions | Files | Quality Gates | Learni
 ```
 [SESSION: role=Lead | task=<desc> | phase=CONTEXT]
 [PHASE: CONTEXT | progress=1/7]
+[SKILLS: loaded=N | available: #1,#2,#3...]
+[KNOWLEDGE: loaded | entities=N | sources=M]
 ```
 
 **During task execution**:
 - [ ] Emit `[SESSION:]` at start ← **FIRST ACTION**
 - [ ] Emit `[PHASE: CONTEXT | progress=1/7]` ← **SECOND ACTION**
+- [ ] Emit `[SKILLS: loaded=N | available: ...]` ← **THIRD ACTION**
+- [ ] Emit `[KNOWLEDGE: loaded | entities=N | sources=M]` ← **FOURTH ACTION**
 - [ ] Emit `[PHASE: PLAN | progress=2/7]` when designing
 - [ ] Emit `[DECISION: ?]` for every choice
+- [ ] Emit `[SKILL: #N Name | applied]` when using a skill
 - [ ] Emit `[ATTEMPT #N]` for every implementation try
 - [ ] Use `#runSubagent` for specialist work → Emit `[SUBAGENT: Name]`
 - [ ] Emit `[PHASE: INTEGRATE | progress=4/7]`

.github/instructions/protocols.md

@@ -7,8 +7,12 @@ applyTo: '**'
 ## Session
 ```
 [SESSION: role=Lead|Architect|Developer|Reviewer|Researcher | task=<desc> | phase=CONTEXT]
+[SKILLS: loaded=N | available: #1,#2,#3...]
+[KNOWLEDGE: loaded | entities=N | sources=M]
 ```
 
+**CRITICAL**: Load and verify knowledge BEFORE proceeding with any work.
+
 ## Delegation
 ```
 [DELEGATE: agent=<specialist> | task=<desc>]
@@ -49,6 +53,13 @@ Artifacts: [files] | Learnings: [patterns]
 [PHASE: CONTEXT|PLAN|COORDINATE|INTEGRATE|VERIFY|LEARN|COMPLETE | progress=N/7 | next=<phase>]
 ```
 
+## Skill Usage
+```
+[SKILL: #N Name | applied] → action/context
+```
+**Purpose**: Transparency - shows which skills are being applied when.
+**Example**: `[SKILL: #3 Security | applied] → Validating input sanitization`
+
 ## Knowledge
 ```
 [KNOWLEDGE: added=N | updated=M | type=project|global]