fix: apply debate findings - context passthrough, canonical redaction, relaxed disagreement (#227)

* fix: remove auto-generated header from adapter files

The `<!-- AUTO-GENERATED ... -->` HTML comment before frontmatter
prevented tools like agnix from parsing YAML frontmatter on line 1.
Removed the header entirely - the adapters/ directory is
self-explanatory. Drops agnix errors from 83 to 8.

* fix: apply debate findings and enhance analysis to consult and debate plugins

Debate skill hardened based on its own first debate's findings:
- Universal evidence standard for both proposer AND challenger
- Proposer prompts now require cited evidence (was challenger-only)
- Challenger follow-up reordered: anti-convergence guard first
- Minimum-disagreement requirement per round added
- Context summarization criteria specified (500-800 tokens)
- Rigor indicator and Debate Quality rating in synthesis output

Consult skill fixes from enhance analysis:
- Gemini section: added missing Session ID extraction line
- Codex: removed invalid -a suggest flag (codex exec doesn't support it)
- Codex: added -c model_reasoning_effort to safe command patterns
- Gemini models: replaced all -preview suffixes with stable names

Also updates README, test strategy doc, and all adapters.

* fix: apply debate findings - context passthrough, canonical redaction, relaxed disagreement

Three fixes from the claude-vs-codex debate findings:
1. --context passthrough from /debate command through orchestrator to
   consult skill, so both tools see the same codebase context
2. Canonical redaction: orchestrator references consult agent's full
   14-pattern table instead of maintaining a drifted 5-pattern subset
3. Relaxed forced disagreement: challenger can now certify a concern
   as genuinely resolved with evidence, not forced to always disagree

* fix: apply debate findings - context passthrough, canonical redaction, relaxed disagreement

Fixes from debate findings + enhance/review analysis:

Debate plugin:
- --model=auto bug: omit --model instead of passing "auto" literally
- --context=file: add follow-up path prompt with validation
- --context added to interactive picker
- Template {round-1} -> {previous_round} for clarity
- Relaxed disagreement: can certify concerns resolved with evidence

Consult plugin:
- Codex model: gpt-5-codex-mini/gpt-5.1-codex-mini -> gpt-5.3-codex-spark
- Redaction table: skill is now canonical source (14 patterns), agent references it
- --continue + --count mutual exclusion in agent
- OpenCode model: claude-opus-4-5 -> claude-opus-4-6

Infrastructure:
- Adapter transform: preserve .claude/ in platform doc enumerations
- New: __tests__/debate-command.test.js (76 tests)
- Updated consult tests for new redaction ownership

* chore: switch exploration-agent and learn-agent from opus to sonnet

New sonnet is faster, cheaper, and close to opus capability. These
agents do pattern matching and synthesis, not deep reasoning:
- exploration-agent: reads code, maps patterns
- learn-agent: web research + guide creation

Keeps opus for: planning, implementation, perf-orchestrator,
debate-orchestrator (judgment-intensive tasks).

* fix: add path containment checks for --context=file in debate command

Author: avifenesh

CLAUDE.md

@@ -183,8 +183,8 @@ agentsys                # Run installer
 
 | Model | Agents | Use Case |
 |-------|--------|----------|
-| **opus** | exploration, planning, implementation, perf-orchestrator | Complex reasoning, analysis |
-| **sonnet** | task-discoverer, delivery-validator, ci-fixer, deslop-agent, reporters | Validation, pattern matching |
+| **opus** | planning, implementation, perf-orchestrator, debate-orchestrator | Complex reasoning, judgment |
+| **sonnet** | exploration, learn, task-discoverer, delivery-validator, ci-fixer, deslop-agent, reporters | Validation, pattern matching |
 | **haiku** | worktree-manager, ci-monitor, simple-fixer | Mechanical execution |
 
 See [README.md](./README.md#command-details) and [docs/reference/AGENTS.md](./docs/reference/AGENTS.md) for full agent list.

__tests__/consult-command.test.js

@@ -324,27 +324,21 @@ describe('security constraints', () => {
     expect(skillContent).toMatch(/delete the temp file/i);
   });
 
-  test('agent has output sanitization with redaction patterns', () => {
+  test('agent references skill as canonical redaction source', () => {
     expect(agentContent).toMatch(/Output Sanitization/);
-    // Key patterns must be present
-    expect(agentContent).toMatch(/sk-\[a-zA-Z0-9/);
-    expect(agentContent).toMatch(/AIza\[a-zA-Z0-9/);
-    expect(agentContent).toMatch(/ghp_/);
-    expect(agentContent).toMatch(/Bearer/);
-    expect(agentContent).toMatch(/AKIA/);
-    expect(agentContent).toMatch(/REDACTED/);
+    // Agent should reference the skill, not maintain its own table
+    expect(agentContent).toMatch(/consult skill|SKILL\.md/i);
   });
 
-  test('agent redaction warning is conditional', () => {
-    expect(agentContent).toMatch(/if any redaction occurs/i);
-  });
-
-  test('skill has inline redaction pattern summary', () => {
-    // Skill should be self-contained with at least key patterns
-    expect(skillContent).toMatch(/sk-.*API key/i);
+  test('skill has full redaction pattern table (canonical source)', () => {
+    // Skill is the single source of truth for redaction patterns
+    expect(skillContent).toMatch(/Output Sanitization/);
+    expect(skillContent).toMatch(/sk-\[a-zA-Z0-9/);
+    expect(skillContent).toMatch(/AIza\[a-zA-Z0-9/);
     expect(skillContent).toMatch(/ghp_/);
     expect(skillContent).toMatch(/Bearer/);
     expect(skillContent).toMatch(/AKIA/);
+    expect(skillContent).toMatch(/REDACTED/);
   });
 });