llm-council

name	llm-council
description	[Decision Support] Use when pressure-testing irreversible, high-stakes decisions with adversarial AI advisors.

Codex compatibility note:

• Invoke repository skills with $skill-name in Codex; this mirrored copy rewrites legacy Claude /skill-name references.
• Prefer the plan-hard skill for planning guidance in this Codex mirror.
• Task tracker mandate: BEFORE executing any workflow or skill step, create/update task tracking for all steps and keep it synchronized as progress changes.
• User-question prompts mean to ask the user directly in Codex.
• Ignore Claude-specific mode-switch instructions when they appear.
• Strict execution contract: when a user explicitly invokes a skill, execute that skill protocol as written.
• Subagent authorization: when a skill is user-invoked or AI-detected and its protocol requires subagents, that skill activation authorizes use of the required spawn_agent subagent(s) for that task.
• Do not skip, reorder, or merge protocol steps unless the user explicitly approves the deviation first.
• For workflow skills, execute each listed child-skill step explicitly and report step-by-step evidence.
• If a required step/tool cannot run in this environment, stop and ask the user before adapting.

Codex Project-Reference Loading (No Hooks)

Codex does not receive Claude hook-based doc injection. When coding, planning, debugging, testing, or reviewing, open project docs explicitly using this routing.

Always read:

• docs/project-config.json (project-specific paths, commands, modules, and workflow/test settings)
• docs/project-reference/docs-index-reference.md (routes to the full docs/project-reference/* catalog)
• docs/project-reference/lessons.md (always-on guardrails and anti-patterns)

Situation-based docs:

• Backend/CQRS/API/domain/entity changes: backend-patterns-reference.md, domain-entities-reference.md, project-structure-reference.md
• Frontend/UI/styling/design-system: frontend-patterns-reference.md, scss-styling-guide.md, design-system/README.md
• Spec/test-case planning or TC mapping: feature-docs-reference.md
• Integration test implementation/review: integration-test-reference.md
• E2E test implementation/review: e2e-test-reference.md
• Code review/audit work: code-review-rules.md plus domain docs above based on changed files

Do not read all docs blindly. Start from docs-index-reference.md, then open only relevant files for the task.

[IMPORTANT] MUST ATTENTION use council only for multi-option, hard-to-reverse, high-stakes decisions. NEVER council trivial, factual, reversible, or single-option questions. [IMPORTANT] MUST ATTENTION spawn 5 advisors in parallel, then 5 fresh peer reviewers in parallel, then chairman synthesis. [IMPORTANT] MUST ATTENTION require evidence for code/architecture claims: file:line, graph trace, or explicit "insufficient evidence."

LLM Council

Quick Summary

Goal: Adversarial decision support for costly wrong choices. Five advisors analyze independently, five fresh reviewers critique anonymously, chairman produces final verdict + report artifacts.

Workflow: Gate → Frame → 5 parallel advisors → anonymized 5-reviewer peer review → chairman synthesis → paired HTML/MD reports.

Key Rules:

• MUST ATTENTION use cheaper ladder first: $why-review → $plan-validate → $llm-council.
• MUST ATTENTION graph-trace code/architecture questions when .code-graph/graph.db exists.
• NEVER let earlier advisor responses bleed into later advisors; parallel spawn required.
• ALWAYS mark verdict degraded if fewer than 5 usable advisor responses return.
• ALWAYS sync edits to .agents/skills/llm-council/SKILL.md.

---

Phase 0: Council Gate

Run before advisors.

Gate	Required	Route if false
Multi-option	>=2 viable paths	Single-option rationale → $why-review
Hard to reverse	Architecture, stack, pricing, hiring, irreversible refactor	Reversible choice → answer directly
Real stakes	Wrong call costs >=1 week, money, trust, or strategic position	Low stakes → answer directly
Multi-angle value	Contrarian/first-principles/upside/outside/execution views change answer	Factual/single-domain → answer directly

If any gate fails, state failed gate + lighter route. NEVER council "should I use markdown."

Good prompts: pricing model, positioning angle, pivot, hiring vs automation, architecture bet, high-risk launch, irreversible refactor. Bad prompts: factual lookup, simple yes/no, content generation, summarization, bugfix, package upgrade, routine refactor, anything decidable with one grep.

---

Advisor Dimensions

Each advisor = thinking dimension, not persona costume. One strong angle each.

Advisor	Think
Contrarian	What fails? What assumption kills plan?
First Principles Thinker	What problem are we solving? Which assumptions need rebuild?
Expansionist	What upside, adjacent opportunity, undervalued path is missing?
Outsider	What confuses someone with no context? What jargon hides value?
Executor	What can happen Monday morning? Fastest validated path?

Tensions: Contrarian vs Expansionist, First Principles vs Executor, Outsider grounds both.

---

Step 1: Frame Question

When user says "council this", enrich then frame.

Context discovery budget: <=30 seconds; read 2-3 high-signal files.

Search order: CLAUDE.md / AGENTS.md; docs/project-config.json; docs/project-reference/project-structure-reference.md; matching docs/project-reference/*{domain-entities,backend-patterns,frontend-patterns,code-review-rules}*; docs/specs/; memory/; user-referenced files; plans/reports/council-*; domain data (pricing -> revenue, architecture -> service map, tech stack -> dependencies).

Code/architecture gate: If question references existing code, services, files, or blast radius, run before framing:

python .claude/scripts/code_graph trace <key-file> --direction both --json

Skip graph only when .code-graph/graph.db missing or question is non-code.

Framed question includes: core decision, user context, workspace evidence, stakes, constraints, known unknowns. Do not add your opinion. Ask exactly one clarifying question only if prompt is too vague.

---

Step 2: Advisor Round

Spawn all 5 advisors simultaneously. Each gets identity, framed question, evidence rules, output constraints.

You are [Advisor Name] on an LLM Council.
Thinking style: [advisor dimension from table]

Question:
---
[framed question]
---

EVIDENCE RULES:
- Code/architecture claims require `file:line`, graph trace, or "I don't have enough evidence yet."
- If existing code context is needed, run:
  python .claude/scripts/code_graph trace <file> --direction both --json
  python .claude/scripts/code_graph connections <file> --json
- Cite trace output for blast radius, callers, downstream impact.
- Confidence: 95-100% full trace | 80-94% main paths | 60-79% partial | <60% do not recommend.
- Do NOT speculate. Name missing evidence instead.

Respond from assigned angle. Direct, specific, unbalanced by design. Other advisors cover other angles.
Length: 150-300 words plus one-line confidence declaration. No preamble.

---

Step 3: Peer Review Round

Collect responses. Randomize/anonymize as Response A-E. Spawn 5 fresh reviewers in parallel.

You are reviewing an LLM Council.

Question:
---
[framed question]
---

Anonymized responses:
**Response A:** [response]
**Response B:** [response]
**Response C:** [response]
**Response D:** [response]
**Response E:** [response]

Answer:
1. Strongest response? Why?
2. Biggest blind spot? What is missing?
3. What did all five miss?

Reference responses by letter. Be specific. Under 200 words.

---

Step 4: Chairman Synthesis

Chairman receives original question, framed question, de-anonymized advisor responses, peer reviews, anonymization map.

You are Chairman of an LLM Council. Synthesize 5 advisors + peer reviews into final verdict.

Question:
---
[framed question]
---

ADVISOR RESPONSES:
**Contrarian:** [response]
**First Principles Thinker:** [response]
**Expansionist:** [response]
**Outsider:** [response]
**Executor:** [response]

PEER REVIEWS:
[all peer reviews]

Produce exact structure:
## Where the Council Agrees
[Independent convergences; high-confidence signals.]
## Where the Council Clashes
[Genuine disagreements. Present both sides; explain why reasonable advisors disagree.]
## Blind Spots the Council Caught
[Only emerged through peer review.]
## The Recommendation
[Clear direct recommendation. Not "it depends."]
## The One Thing to Do First
[Single concrete next step.]

Be direct. Do not hedge. Give clarity unavailable from one perspective.

Quality loop: If chairman misses required sections, ignores degraded-quality state, or makes unsupported code claims, repair with fresh chairman prompt. Max 3 repair rounds; then escalate with missing evidence.

---

Step 5: Report Artifacts

Write both files; create plans/reports/ if missing.

plans/reports/council-{YYMMDD-HHMM}-{kebab-slug}.html
plans/reports/council-{YYMMDD-HHMM}-{kebab-slug}.md

{YYMMDD-HHMM} = session datetime. {kebab-slug} = 3-6 word question descriptor. Both share prefix. NEVER write artifacts to workspace root or docs/.

HTML: self-contained inline CSS, question top, prominent chairman verdict, agreement/disagreement visual, collapsed advisor responses, collapsed peer review highlights, footer timestamp/topic, professional briefing style, open after generation.

Markdown transcript: original question, framed question, all advisor responses, all peer reviews, anonymization mapping, chairman synthesis.

---

Output Format

Council complete.

Report: plans/reports/council-{YYMMDD-HHMM}-{kebab-slug}.html
Transcript: plans/reports/council-{YYMMDD-HHMM}-{kebab-slug}.md

Verdict: [1-2 sentence chairman recommendation]
First action: [single next step]

---

Workflow Integration

Opt-in escalation hook from host skills. NEVER wire into bugfix, refactor, migration, package-upgrade, or test-* workflows.

Host skill	Mode	Default	Gate
architecture-design	Always-offer after ## Next Steps	Skip	User chooses
tech-stack-research	Always-offer after ## Next Steps	Skip	User chooses
why-review	Conditional on active plan/PBI frontmatter	Escalate when gate fires	8-OR gate
prioritize	Conditional on ranking output	Escalate when gate fires	RICE top-2 within 15%, MoSCoW tie, or stakeholder disagreement

why-review Gate Schema

Gate fires when ANY field true. Absent fields default no-fire; gate opt-in via frontmatter, never opt-out.

Field	Convention	Fires when
cross_service_impact	NONE / PARTIAL / FULL	value != NONE
breaking_changes	bool	true
complexity	low / medium / high / critical	high, critical, or story_points >= 13
new_framework	bool	true
irreversible	bool	true
security_critical	bool	true
performance_critical	bool	true
cost_high	bool	true

Host prompt copy MUST cite cheaper rungs: $why-review, $plan-validate, $llm-council.

---

Closing Reminders

IMPORTANT MUST ATTENTION use council only for multi-option, hard-to-reverse, high-stakes decisions. IMPORTANT MUST ATTENTION spawn 5 advisors in parallel, then 5 fresh peer reviewers in parallel, then chairman synthesis. IMPORTANT MUST ATTENTION require evidence for code/architecture claims: file:line, graph trace, or explicit "insufficient evidence." IMPORTANT MUST ATTENTION mark verdict degraded if fewer than 5 usable advisor responses return. IMPORTANT MUST ATTENTION write paired HTML + Markdown artifacts under plans/reports/ and open HTML. IMPORTANT MUST ATTENTION sync every canonical edit to .agents/skills/llm-council/SKILL.md.

Anti-Rationalization:

Evasion	Rebuttal
"This decision feels important"	Gate it: multi-option, hard-to-reverse, real stakes, multi-angle value.
"One advisor can handle it"	Council value comes from independent angles + anonymous peer review.
"Sequential spawn is simpler"	Sequential spawn contaminates independence. Parallel spawn required.
"Four advisors is close enough"	Missing angle changes verdict quality. Mark degraded.
"Evidence would slow us down"	Unsupported code/architecture claims are speculation. Use graph/file proof or say insufficient evidence.

Hookless Prompt Protocol Mirror (Auto-Synced)

Source: .claude/hooks/lib/prompt-injections.cjs + .claude/.ck.json

[WORKFLOW-EXECUTION-PROTOCOL] [BLOCKING] Workflow Execution Protocol — MANDATORY IMPORTANT MUST CRITICAL. Do not skip for any reason.

1. DETECT: Match prompt against workflow catalog
2. ANALYZE: Find best-match workflow AND evaluate if a custom step combination would fit better
3. ASK (REQUIRED FORMAT): Use a direct user question with this structure:
   • Question: "Which workflow do you want to activate?"
   • Option 1: "Activate [BestMatch Workflow] (Recommended)"
   • Option 2: "Activate custom workflow: [step1 → step2 → ...]" (include one-line rationale)
4. ACTIVATE (if confirmed): Call $workflow-start <workflowId> for standard; sequence custom steps manually
5. CREATE TASKS: task tracking for ALL workflow steps
6. EXECUTE: Follow each step in sequence [CRITICAL-THINKING-MINDSET] Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence >80% to act. Anti-hallucination principle: Never present guess as fact — cite sources for every claim, admit uncertainty freely, self-check output for errors, cross-reference independently, stay skeptical of own confidence — certainty without evidence root of all hallucination. AI Attention principle (Primacy-Recency): Put the 3 most critical rules at both top and bottom of long prompts/protocols so instruction adherence survives long context windows.

Learned Lessons

Lessons Learned

[CRITICAL] Hard-won project debugging/architecture rules. MUST ATTENTION apply BEFORE forming hypothesis or writing code.

Quick Summary

Goal: Prevent recurrence of known failure patterns — debugging, architecture, naming, AI orchestration, environment.

Top Rules (apply always):

• MUST ATTENTION verify ALL preconditions (config, env, DB names, DI regs) BEFORE code-layer hypothesis
• MUST ATTENTION fix responsible layer — NEVER patch symptom sites with caller-specific defensive code
• MUST ATTENTION use ExecuteInjectScopedAsync for parallel async + repo/UoW — NEVER ExecuteUowTask
• MUST ATTENTION name by PURPOSE not CONTENT — adding member forces rename = abstraction broken
• MUST ATTENTION persist sub-agent findings incrementally after each file — NEVER batch at end
• MUST ATTENTION Windows bash: verify Python alias (where python/where py) — NEVER assume python/python3 resolves

---

Debugging & Root Cause Reasoning

• [2026-04-11] Holistic-first: verify environment before code. Failure → list ALL preconditions (config, env vars, DB names, endpoints, DI regs, credentials, permissions, data prerequisites) → verify each via evidence (grep/cat/query) BEFORE code-layer hypothesis. Worst rabbit holes: diving nearest layer while bug sits elsewhere — e.g., hours debugging "sync timeout", real cause: test appsettings pointing wrong DB. Cheapest check first.
• [2026-04-01] Ask "whose responsibility?" before fixing. Trace: bug in caller (wrong data) or callee (wrong handling)? Fix responsible layer — NEVER patch symptom site masking real issue.
• [2026-04-01] Trace data lifecycle, not error site. Follow data: creation → transformation → consumption. Bug usually where data created wrong, not consumed.
• [2026-04-01] Code is caller-agnostic. Functions/handlers/consumers don't know who invokes them. Comments/guards/messages describe business intent — NEVER reference specific callers (tests, seeders, scripts).

Architecture Invariants

• [2026-03-31] ParallelAsync + repo/UoW MUST use ExecuteInjectScopedAsync, NEVER ExecuteUowTask. ExecuteUowTask creates new UoW but reuses outer DI scope (same DbContext) — parallel iterations sharing non-thread-safe DbContext silently corrupt data. ExecuteInjectScopedAsync creates new UoW + new DI scope (fresh repo per iteration).
• [2026-03-31] Bus message naming MUST include service name prefix — core services NEVER consume feature events. Prefix declares schema ownership (AccountUserEntityEventBusMessage = Accounts owns). Core services (Accounts, Communication) are leaders. Feature services (Growth, Talents) sending to core MUST use {CoreServiceName}...RequestBusMessage — never define own event for core to consume.

Naming & Abstraction

• [2026-04-12] Name PURPOSE not CONTENT — "OrXxx" anti-pattern. HrManagerOrHrOrPayrollHrOperationsPolicy names set members, not what it guards. Add role → rename = broken abstraction. Rule: names express DOES/GUARDS, not CONTAINS. Test: adding/removing member forces rename? YES = content-driven = bad → rename to purpose (e.g., HrOperationsAccessPolicy). Nuance: "Or" fine in behavioral idioms (FirstOrDefault, SuccessOrThrow) — expresses HAPPENS, not membership.

Environment & Tooling

• [2026-04-20] Windows bash: NEVER assume python/python3 resolves — verify alias first. Python may not be in bash PATH under those names. Check: where python / where py. Prefer py (Windows Python Launcher) for one-liners, node if JS alternative exists.

Test-specific lessons → docs/project-reference/integration-test-reference.md Lessons Learned section. Production-code anti-patterns → docs/project-reference/backend-patterns-reference.md Anti-Patterns section. Generic debugging/refactoring reminders → System Lessons in .claude/hooks/lib/prompt-injections.cjs.

---

Closing Reminders

• IMPORTANT MUST ATTENTION holistic-first: verify ALL preconditions (config, env, DB names, endpoints, DI regs) BEFORE code-layer hypothesis — cheapest check first
• IMPORTANT MUST ATTENTION fix responsible layer — NEVER patch symptom site; trace caller (wrong data) vs callee (wrong handling), fix root owner
• IMPORTANT MUST ATTENTION parallel async + repo/UoW → ALWAYS ExecuteInjectScopedAsync, NEVER ExecuteUowTask (shared DbContext = silent data corruption)
• IMPORTANT MUST ATTENTION bus message prefix = schema ownership; feature services NEVER define events for core services — use {CoreServiceName}...RequestBusMessage
• IMPORTANT MUST ATTENTION name by PURPOSE — adding/removing member forces rename = broken abstraction
• IMPORTANT MUST ATTENTION sub-agents MUST write findings after each file/section — NEVER batch all findings into one final write
• IMPORTANT MUST ATTENTION Windows bash: NEVER assume python/python3 resolves — run where python/where py first, use py launcher or node

[LESSON-LEARNED-REMINDER] [BLOCKING] Task Planning & Continuous Improvement — MANDATORY. Do not skip.

Break work into small tasks (task tracking) before starting. Add final task: "Analyze AI mistakes & lessons learned".

Extract lessons — ROOT CAUSE ONLY, not symptom fixes:

1. Name the FAILURE MODE (reasoning/assumption failure), not symptom — "assumed API existed without reading source" not "used wrong enum value".
2. Generality test: does this failure mode apply to ≥3 contexts/codebases? If not, abstract one level up.
3. Write as a universal rule — strip project-specific names/paths/classes. Useful on any codebase.
4. Consolidate: multiple mistakes sharing one failure mode → ONE lesson.
5. Recurrence gate: "Would this recur in future session WITHOUT this reminder?" — No → skip $learn.
6. Auto-fix gate: "Could $code-review/$code-simplifier/$security/$lint catch this?" — Yes → improve review skill instead.
7. BOTH gates pass → ask user to run $learn. [TASK-PLANNING] [MANDATORY] BEFORE executing any workflow or skill step, create/update task tracking for all planned steps, then keep it synchronized as each step starts/completes.