Agent Skills: Lessons Learned

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.

[FINAL PURPOSE REMINDER — MUST ATTENTION CRITICAL]

Ensure the changes is reasonable, no potential bugs or flaws, critical thinking hard.

[BLOCKING] Execute skill steps in declared order. NEVER skip, reorder, or merge steps without explicit user approval. [BLOCKING] Before each step or sub-skill call, update task tracking: set in_progress when step starts, set completed when step ends. [BLOCKING] Every completed/skipped step MUST include brief evidence or explicit skip reason. [BLOCKING] If Task tools are unavailable, create and maintain an equivalent step-by-step plan tracker with the same status transitions.

Quick Summary

Goal: Validate that a plan contains sufficient design rationale (WHY, not just WHAT) before implementation begins.

Applies to: Features and refactors only — bugfixes and trivial changes exempt.

Why this exists: AI code generation optimizes mechanics but misses conceptual quality. This skill ensures the human thinking happened before the mechanical coding starts.

Your Mission

Adversarial Review Mindset (NON-NEGOTIABLE)

Default stance: SKEPTIC, not validator. Your job is to find what's wrong, not confirm what's right.

Confirmation bias trap: After reading a coherent plan, AI naturally finds reasons to agree. The current context (post-plan, post-fix) amplifies this — you've already seen the reasoning and rationalized it. This section exists to break that loop.

Adversarial Techniques (apply ALL before concluding)

1. Steel-Man Protocol Before dismissing any rejected alternative, argue FOR it as strongly as possible. Ask: "Would a senior engineer with 10 years of experience in this domain actually choose this alternative?" If yes — the plan's dismissal needs stronger justification.

2. Why NOT? Inversion For every stated reason ("We chose X because Y"), ask: "Why NOT X? What does X sacrifice?" The plan must acknowledge what the chosen approach gives up, not just what it gains.

3. What If? Assumption Stress Test List the top 3 assumptions in the plan. For each: "What if this assumption is wrong?" A good plan survives at least 2 of its 3 assumptions being false.

4. Pre-Mortem Assume the plan ships and fails in production within 3 months. Write one concrete failure scenario that is plausible given the current plan. If you can't find one, you haven't looked hard enough.

5. Unseen Alternatives Identify 1-2 approaches NOT mentioned in the plan at all. Did the author genuinely not consider them, or did they consider and consciously exclude them? Missing alternatives without exclusion reasoning = weak coverage.

6. Pros/Cons Symmetry Check Count the pros listed for the chosen approach. Count the cons. If pros > cons by more than 2:1, the analysis is likely biased. Real trade-offs have roughly equal weight on both sides.

7. Contrarian Pass Before writing any finding, generate at least 2 sentences arguing the OPPOSITE conclusion. If you're about to write PASS — argue for NEEDS WORK. If about to write NEEDS WORK — argue for PASS. Then decide which argument is stronger.

Forbidden Patterns

• Leading with confirmation: "This looks good because..." → STOP. Lead with challenges first.
• Presence = quality: "Alternatives section exists ✅" → presence is NOT quality. Were they real alternatives?
• Vague rationale: "We chose X because it's better" → What metric? Better at what cost?
• Asymmetric trade-offs: Listing 3 pros and 1 con → the analysis is likely incomplete.
• "Looks fine" without evidence of adversarial challenge

Anti-Bias Gate (MANDATORY before finalizing verdict)

Complete ALL checks before writing the final verdict:

• MUST ATTENTION steel-man at least one rejected alternative (argue FOR it)
• MUST ATTENTION identify at least 1 alternative NOT in the plan
• MUST ATTENTION list 2-3 arguments AGAINST the chosen approach
• MUST ATTENTION surface 2-3 hidden assumptions with stress tests
• MUST ATTENTION run the pre-mortem (one concrete failure scenario)
• MUST ATTENTION check pros/cons symmetry

If any check is incomplete → you have NOT completed the adversarial review. Go back.

Behavioral Delta Matrix (MANDATORY for bugfixes)

Plan Resolution

1. If arguments contain a path → use that plan directory
2. Else check ## Plan Context in injected context → use active plan path
3. If no plan found → tell user: "No active plan found. Run $plan-hard or $plan-hard first."

Validation Checklist

Read the plan's plan.md and all phase-*.md files. Check each item below. Two dimensions per check: presence AND quality depth.

Rule: Presence alone is NOT a pass. A section that exists but contains weak, asymmetric, or unverified reasoning FAILS quality depth.

Required Sections (in plan.md or phase files)

| # | Section | Presence Check | Quality Depth Check (adversarial) | | --- | --------------------------- | ------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | 1 | Problem Statement | 2-3 sentences describing the problem | Is the problem scoped correctly? Could it be framed differently to lead to a different solution? Are symptoms confused with root cause? | | 2 | Alternatives Considered | Minimum 2 alternatives listed with pros/cons | Are alternatives real (not strawmen)? Would a domain expert seriously consider each? Are the cons of the CHOSEN approach listed, not just cons of the others? | | 3 | Design Rationale | Explicit reasoning linking decision to trade-offs | Is reasoning causal (X leads to Y) or just descriptive (X is better)? Are hidden assumptions surfaced? Does it address failure modes, not just success modes? | | 4 | Risk Assessment | At least 1 risk per phase | Are risks ranked by severity? Are mitigations concrete actions or vague intentions ("monitor closely")? Is there at least one risk about the approach itself (not just execution)? | | 5 | Ownership | Clear who maintains code post-merge | Implicit OK (author owns), explicit better |

Optional (Flag if Missing, Don't Fail)

| # | Section | When Required | Quality Depth Check | | --- | ------------------------ | --------------------------------------- | ------------------------------------------------------------------ | | 6 | Operational Impact | Service-layer or API changes | Are rollback steps defined? What breaks if this is reverted? | | 7 | Cross-Service Impact | Changes touching multiple microservices | Are all downstream consumers identified? Who needs to be notified? | | 8 | Migration Strategy | Database schema or data changes | Is there a rollback plan? Is it tested on a data sample? |

Output Format

## Why-Review Results

**Plan:** {plan path}
**Date:** {date}
**Verdict:** PASS / NEEDS WORK

### Checklist

| #   | Check                   | Presence | Quality Depth | Notes                            |
| --- | ----------------------- | -------- | ------------- | -------------------------------- |
| 1   | Problem Statement       | ✅/❌    | ✅/⚠️/❌      | {what's strong / what's weak}    |
| 2   | Alternatives Considered | ✅/❌    | ✅/⚠️/❌      | {are they real or strawmen?}     |
| 3   | Design Rationale        | ✅/❌    | ✅/⚠️/❌      | {causal or just descriptive?}    |
| 4   | Risk Assessment         | ✅/❌    | ✅/⚠️/❌      | {concrete mitigations or vague?} |
| 5   | Ownership               | ✅/❌    | ✅/⚠️/❌      | {details}                        |

> ✅ Strong ⚠️ Weak/Partial ❌ Missing

### Adversarial Analysis

**Strongest arguments AGAINST the chosen approach:**

1. {argument 1 — cite specific plan text that weakens under this pressure}
2. {argument 2}
3. {argument 3 if applicable}

**Unexamined alternatives** (not mentioned in the plan):

- {alternative A} — why it might be worth considering
- {alternative B if applicable}

**Weakest assumptions** (if wrong, the plan breaks):

1. {assumption} — impact if false: {consequence}
2. {assumption} — impact if false: {consequence}

**Pre-mortem** (assume it ships and fails in 3 months):

> {One concrete, plausible failure scenario based on the plan's approach}

**Pros/Cons symmetry:** Pros listed: {N} | Cons listed: {N} | Bias: {balanced / leans toward pros / leans toward cons}

### Missing Items (if any)

- {specific item to add before implementation}

### Recommendation

{Proceed to $cook | Add missing sections first | Add adversarial analysis to plan before proceeding}

Round 2: Adversarial Re-Review (MANDATORY)

Protocol: Deep Multi-Round Review (inlined via SYNC:double-round-trip-review above)

After completing Round 1 checklist evaluation, execute a second full review round using adversarial mode:

1. Assume Round 1 was wrong — start with: "Round 1 missed something. Find it."
2. Challenge every PASS item from Round 1 — generate at least 2 sentences arguing the opposite for each
3. Complete the Anti-Bias Gate (all 6 boxes from Adversarial Review Mindset section)
4. Populate the Adversarial Analysis section in the output — this is MANDATORY:
   • At least 2 arguments against the chosen approach
   • At least 1 unexamined alternative
   • At least 2 hidden assumptions with failure consequences
   • Pre-mortem scenario
   • Pros/Cons symmetry count
5. Focus on what Round 1 typically misses:
   • Alternatives that are strawmen (too easy to dismiss)
   • Risks stated vaguely without concrete mitigations
   • Assumptions embedded in the problem statement itself
   • Scope creep disguised as "related improvements"
6. Update verdict if Round 2 found new issues
7. Final verdict must incorporate findings from BOTH rounds AND the Adversarial Analysis

Scope

• Applies to: Features, refactors, architectural changes
• Exempt: Bugfixes, config changes, single-file tweaks, documentation-only
• Enforcement: Advisory (soft warning) — does not block implementation

Important Notes

• Review only — do NOT modify plan files or implement changes
• Keep output concise — actionable in <2 minutes
• Simple plans still require the Anti-Bias Gate — findings may be brief ("No real alternatives found — approach is the only viable path given X constraint"), but the gate cannot be skipped

Next Steps

MANDATORY IMPORTANT MUST ATTENTION — NO EXCEPTIONS after completing this skill, you MUST ATTENTION use a direct user question to present these options. Do NOT skip because the task seems "simple" or "obvious" — the user decides:

• "$cook (Recommended)" — Begin implementation after design rationale is validated
• "$code" — If implementing a simpler change
• "Skip, continue manually" — user decides

Additionally — conditional $llm-council escalation

After the existing ## Next Steps ask the user directly call completes, evaluate the gate:

Step A — Workflow blacklist suppression (run FIRST, short-circuits):

The host llm-council skill blacklists routine workflows where 11 sub-agent calls are not worth the decision impact. If $why-review is invoked inside ANY of the following workflows, suppress the council gate entirely and skip Step B:

• refactor (behavior-preserving by definition — reversibility is high)
• migration (schema/data migration is procedural, not a multi-option strategic choice)
• package-upgrade (version bumps — irreversibility caught upstream by $tech-stack-research when major)
• performance (optimization with measurable rollback)
• verification (validation-only, not a new commit)
• bugfix (exempt from $why-review anyway — listed for clarity)
• Any workflow whose ID starts with test-* (test-only)

Detect workflow context by reading plans/.workflow-state.json (or equivalent active-workflow marker) for the workflowId field. If the file is missing or the workflowId is not in the blacklist above, proceed to Step B.

Rationale: Council cost is 11 LLM calls. The 8-OR frontmatter gate can fire on these workflows (e.g., irreversible: true on a non-trivial refactor) but the decision impact does not justify the cost — $why-review itself is already the right rung. This suppression matches the canonical blacklist in .claude/skills/llm-council/SKILL.md "Workflow Integration" section.

Step B — Frontmatter gate evaluation:

Read the active plan's plan.md frontmatter (or PBI artifact frontmatter when invoked outside a plan). Gate fires when ANY of these 8 conditions evaluates true (per Phase 01 design contract §2 Gate Evaluation Procedure):

1. cross_service_impact ≠ NONE (value is PARTIAL or FULL; case-insensitive)
2. breaking_changes === true (lenient boolean coercion: true, 'true', True fire; false/missing do NOT)
3. complexity ∈ {high, critical} (case-insensitive) OR story_points >= 13 (numeric, after string→int coercion)
4. new_framework === true
5. irreversible === true
6. security_critical === true
7. performance_critical === true
8. cost_high === true

Absent fields default to no-fire — the gate is opt-in via frontmatter, never opt-out.

Plan-level override: Frontmatter may include council_suppress: true to opt-out even when Step A doesn't suppress (e.g., a refactor PBI that was migrated into the feature workflow but the team has already councilled the decision). When set, skip the council prompt and log the suppression reason in ## Next Steps output.

If Step A suppressed OR Step B gate does NOT fire → do NOT mention $llm-council (avoids cost normalization). The skill ends after the existing ## Next Steps prompt.

If Step A did NOT suppress AND Step B gate fires, immediately follow with a SECOND a direct user question invocation (separate from the existing ## Next Steps call — do NOT merge bullet lists):

• "Escalate to $llm-council (Recommended)" — Gate fired (high-stakes signal detected). Run 11 sub-agent council (5 advisors + 5 reviewers + chairman). Use when $why-review alone is insufficient. Cheaper alternatives already exhausted at this point: $plan-validate is the prior rung.
• "Skip — proceed without council" — Acknowledge the gate; proceed with current decision anyway.

[BLOCKING] This is a validation gate. MUST ATTENTION use a direct user question to present review findings and get user confirmation. Completing without asking at least one question is a violation.

[IMPORTANT] Use task tracking to break ALL work into small tasks BEFORE starting — including tasks for each file read. This prevents context loss from long files. For simple tasks, AI MUST ATTENTION ask user whether to skip.

Critical Purpose: Ensure quality — no flaws, no bugs, no missing updates, no stale content. Verify both code AND documentation.

External Memory: For complex or lengthy work (research, analysis, scan, review), write intermediate findings and final results to a report file in plans/reports/ — prevents context loss and serves as deliverable.

Evidence Gate: MANDATORY IMPORTANT MUST ATTENTION — every claim, finding, and recommendation requires file:line proof or traced evidence with confidence percentage (>80% to act, <80% must verify first).

OOP & DRY Enforcement: MANDATORY IMPORTANT MUST ATTENTION — flag duplicated patterns that should be extracted to a base class, generic, or helper. Classes in the same group or suffix (ex *Entity, *Dto, *Service, etc...) MUST ATTENTION inherit a common base (even if empty now — enables future shared logic and child overrides). Verify project has code linting/analyzer configured for the stack.

Behavioral Delta Matrix — MANDATORY for bugfix reviews. Produce this table BEFORE PASS/FAIL verdict. Narrative descriptions don't substitute.

| Input state | Pre-fix behavior | Post-fix behavior | Delta | | ----------- | ------------------ | ----------------- | ------------------------------------ | | {condition} | {current behavior} | {fixed behavior} | Preserved ✓ / Fixed ✓ / REGRESSION ✗ |

Rules: ≥3 rows · ≥1 row the bug report did NOT mention · REGRESSION delta → FAIL until a preservation test covers it (tdd-spec-template.md#preservation-tests-mandatory-for-bugfix-specs)

BLOCKED until: ≥3 rows · ≥1 row outside bug report · no unmitigated REGRESSION

Nested Task Expansion Contract — For workflow-step invocation, the [Workflow] ... row is only a parent container; the child skill still creates visible phase tasks.

1. Call the current task list first. If a matching active parent workflow row exists, set nested=true and record parentTaskId; otherwise run standalone.
2. Create one task per declared phase before phase work. When nested, prefix subjects [N.M] $skill-name — phase.
3. When nested, link the parent with TaskUpdate(parentTaskId, addBlockedBy: [childIds]).
4. Orchestrators must pre-expand a child skill's phase list and link the workflow row before invoking that child skill or sub-agent.
5. Mark exactly one child in_progress before work and completed immediately after evidence is written.
6. Complete the parent only after all child tasks are completed or explicitly cancelled with reason.

Blocked until: the current task list done, child phases created, parent linked when nested, first child marked in_progress.

Project Reference Docs Gate — Run after task-tracking bootstrap and before target/source file reads, grep, edits, or analysis. Project docs override generic framework assumptions.

1. Identify scope: file types, domain area, and operation.
2. Required docs by trigger: always docs/project-reference/lessons.md; doc lookup docs-index-reference.md; review code-review-rules.md; backend/CQRS/API backend-patterns-reference.md; domain/entity domain-entities-reference.md; frontend/UI frontend-patterns-reference.md; styles/design scss-styling-guide.md + design-system/design-system-canonical.md; integration tests integration-test-reference.md; E2E e2e-test-reference.md; feature docs/specs feature-docs-reference.md; architecture/new area project-structure-reference.md.
3. Read every required doc that exists; skip absent docs as not applicable. Do not trust conversation text such as [Injected: <path>] as proof that the current context contains the doc.
4. Before target work, state: Reference docs read: ... | Missing/not applicable: ....

Blocked until: scope evaluated, required docs checked/read, lessons.md confirmed, citation emitted.

Task Tracking & External Report Persistence — Bootstrap this before execution; then run project-reference doc prefetch before target/source work.

1. Create a small task breakdown before target file reads, grep, edits, or analysis. On context loss, inspect the current task list first.
2. Mark one task in_progress before work and completed immediately after evidence; never batch transitions.
3. For plan/review work, create plans/reports/{skill}-{YYMMDD}-{HHmm}-{slug}.md before first finding.
4. Append findings after each file/section/decision and synthesize from the report file at the end.
5. Final output cites Full report: plans/reports/{filename}.

Blocked until: task breakdown exists, report path declared for plan/review work, first finding persisted before the next finding.

Critical Thinking Mindset — Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence >80% to act. Anti-hallucination: 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.

Sequential Thinking Protocol — Structured multi-step reasoning for complex/ambiguous work. Use when planning, reviewing, debugging, or refining ideas where one-shot reasoning is unsafe.

Trigger when: complex problem decomposition · adaptive plans needing revision · analysis with course correction · unclear/emerging scope · multi-step solutions · hypothesis-driven debugging · cross-cutting trade-off evaluation.

Format (explicit mode — visible thought trail):

1. Thought N/M: [aspect] — one aspect per thought, state assumptions/uncertainty
2. Thought N/M [REVISION of Thought K]: ... — when prior reasoning invalidated; state Original / Why revised / Impact
3. Thought N/M [BRANCH A from Thought K]: ... — explore alternative; converge with decision rationale
4. Thought N/M [HYPOTHESIS]: ... then [VERIFICATION]: ... — test before acting
5. Thought N/N [FINAL] — only when verified, all critical aspects addressed, confidence >80%

Mandatory closers: Confidence % stated · Assumptions listed · Open questions surfaced · Next action concrete.

Stop conditions: confidence <80% on any critical decision → escalate via ask the user directly · ≥3 revisions on same thought → re-frame the problem · branch count >3 → split into sub-task.

Implicit mode: apply methodology internally without visible markers when adding markers would clutter the response (routine work where reasoning aids accuracy).

Deep-dive: see $sequential-thinking skill (.claude/skills/sequential-thinking/SKILL.md) for worked examples (api-design, debug, architecture), advanced techniques (spiral refinement, hypothesis testing, convergence), and meta-strategies (uncertainty handling, revision cascades).

AI Mistake Prevention — Failure modes to avoid on every task:

• Check downstream references before deleting. Deleting components causes documentation and code staleness cascades. Map all referencing files before removal.
• Verify AI-generated content against actual code. AI hallucinates APIs, class names, and method signatures. Always grep to confirm existence before documenting or referencing.
• Trace full dependency chain after edits. Changing a definition misses downstream variables and consumers derived from it. Always trace the full chain.
• Trace ALL code paths when verifying correctness. Confirming code exists is not confirming it executes. Always trace early exits, error branches, and conditional skips — not just happy path.
• When debugging, ask "whose responsibility?" before fixing. Trace whether bug is in caller (wrong data) or callee (wrong handling). Fix at responsible layer — never patch symptom site.
• Assume existing values are intentional — ask WHY before changing. Before changing any constant, limit, flag, or pattern: read comments, check git blame, examine surrounding code.
• Verify ALL affected outputs, not just the first. Changes touching multiple stacks require verifying EVERY output. One green check is not all green checks.
• Holistic-first debugging — resist nearest-attention trap. When investigating any failure, list EVERY precondition first (config, env vars, DB names, endpoints, DI registrations, data preconditions), then verify each against evidence before forming any code-layer hypothesis.
• Surgical changes — apply the diff test. Bug fix: every changed line must trace directly to the bug. Don't restyle or improve adjacent code. Enhancement task: implement improvements AND announce them explicitly.
• Surface ambiguity before coding — don't pick silently. If request has multiple interpretations, present each with effort estimate and ask. Never assume all-records, file-based, or more complex path.

Evidence-Based Reasoning — Speculation is FORBIDDEN. Every claim needs proof.

1. Cite file:line, grep results, or framework docs for EVERY claim
2. Declare confidence: >80% act freely, 60-80% verify first, <60% DO NOT recommend
3. Cross-service validation required for architectural changes
4. "I don't have enough evidence" is valid and expected output

BLOCKED until: - [ ] Evidence file path (file:line) - [ ] Grep search performed - [ ] 3+ similar patterns found - [ ] Confidence level stated

Forbidden without proof: "obviously", "I think", "should be", "probably", "this is because" If incomplete → output: "Insufficient evidence. Verified: [...]. Not verified: [...]."

Fix-Triggered Re-Review Loop — Re-review is triggered by a FIX CYCLE, not by a round number. Review purpose: review → if issues → fix → re-review until a round finds no issues. A clean review ENDS the loop — no further rounds required.

Round 1: Main-session review. Read target files, build understanding, note issues. Output findings + verdict (PASS / FAIL).

Decision after Round 1:

• No issues found (PASS, zero findings) → review ENDS. Do NOT spawn a fresh sub-agent for confirmation.
• Issues found (FAIL, or any non-zero findings) → fix the issues, then spawn a fresh sub-agent for Round 2 re-review.

Fresh sub-agent re-review (after every fix cycle): Spawn a NEW spawn_agent tool call — never reuse a prior agent. Sub-agent re-reads ALL files from scratch with ZERO memory of prior rounds. See SYNC:fresh-context-review for the spawn mechanism and SYNC:review-protocol-injection for the canonical Agent prompt template. Each fresh round must catch:

• Cross-cutting concerns missed in the prior round
• Interaction bugs between changed files
• Convention drift (new code vs existing patterns)
• Missing pieces that should exist but don't
• Subtle edge cases the prior round rationalized away
• Regressions introduced by the fixes themselves

Loop termination: After each fresh round, repeat the same decision: clean → END; issues → fix → next fresh round. Continue until a round finds zero issues, or 3 fresh-subagent rounds max, then escalate to user via a direct user question.

Rules:

• A clean Round 1 ENDS the review — no mandatory Round 2
• NEVER skip the fresh sub-agent re-review after a fix cycle (every fix invalidates the prior verdict)
• NEVER reuse a sub-agent across rounds — every iteration spawns a NEW Agent call
• Main agent READS sub-agent reports but MUST NOT filter, reinterpret, or override findings
• Max 3 fresh-subagent rounds per review — if still FAIL, escalate via a direct user question (do NOT silently loop)
• Track round count in conversation context (session-scoped)
• Final verdict must incorporate ALL rounds executed

Report must include ## Round N Findings (Fresh Sub-Agent) for every round N≥2 that was executed.

Fresh Sub-Agent Review — Eliminate orchestrator confirmation bias via isolated sub-agents.

Why: The main agent knows what it (or $cook) just fixed and rationalizes findings accordingly. A fresh sub-agent has ZERO memory, re-reads from scratch, and catches what the main agent dismissed. Sub-agent bias is mitigated by (1) fresh context, (2) verbatim protocol injection, (3) main agent not filtering the report.

When: ONLY after a fix cycle. A review round that finds zero issues ENDS the loop — do NOT spawn a confirmation sub-agent. A review round that finds issues triggers: fix → fresh sub-agent re-review.

How:

1. Spawn a NEW spawn_agent tool call — use code-reviewer subagent_type for code reviews, general-purpose for plan/doc/artifact reviews
2. Inject ALL required review protocols VERBATIM into the prompt — see SYNC:review-protocol-injection for the full list and template. Never reference protocols by file path; AI compliance drops behind file-read indirection (see SYNC:shared-protocol-duplication-policy)
3. Sub-agent re-reads ALL target files from scratch via its own tool calls — never pass file contents inline in the prompt
4. Sub-agent writes structured report to plans/reports/{review-type}-round{N}-{date}.md
5. Main agent reads the report, integrates findings into its own report, DOES NOT override or filter

Rules:

• SKIP fresh sub-agent when the prior round found zero issues (no fixes = nothing new to verify)
• NEVER skip fresh sub-agent after a fix cycle — every fix invalidates the prior verdict
• NEVER reuse a sub-agent across rounds — every fresh round spawns a NEW spawn_agent call
• Max 3 fresh-subagent rounds per review — escalate via a direct user question if still failing; do NOT silently loop or fall back to any prior protocol
• Track iteration count in conversation context (session-scoped, no persistent files)

Review Protocol Injection — Every fresh sub-agent review prompt MUST embed 10 protocol blocks VERBATIM. The template below has ALL 10 bodies already expanded inline. Copy the template wholesale into the Agent call's prompt field at runtime, replacing only the {placeholders} in Task / Round / Reference Docs / Target Files / Output sections with context-specific values. Do NOT touch the embedded protocol sections.

Why inline expansion: Placeholder markers would force file-read indirection at runtime. AI compliance drops significantly behind indirection (see SYNC:shared-protocol-duplication-policy). Therefore the template carries all 10 protocol bodies pre-embedded.

Subagent Type Selection

• code-reviewer — for code reviews (reviewing source files, git diffs, implementation)
• general-purpose — for plan / doc / artifact reviews (reviewing markdown plans, docs, specs)

Canonical Agent Call Template (Copy Verbatim)

spawn_agent({
  description: "Fresh Round {N} review",
  agent_type: "code-reviewer",
  prompt: `
## Task
{review-specific task — e.g., "Review all uncommitted changes for code quality" | "Review plan files under {plan-dir}" | "Review integration tests in {path}"}

## Round
Round {N}. You have ZERO memory of prior rounds. Re-read all target files from scratch via your own tool calls. Do NOT trust anything from the main agent beyond this prompt.

## Protocols (follow VERBATIM — these are non-negotiable)

### Evidence-Based Reasoning
Speculation is FORBIDDEN. Every claim needs proof.
1. Cite file:line, grep results, or framework docs for EVERY claim
2. Declare confidence: >80% act freely, 60-80% verify first, <60% DO NOT recommend
3. Cross-service validation required for architectural changes
4. "I don't have enough evidence" is valid and expected output
BLOCKED until: Evidence file path (file:line) provided; Grep search performed; 3+ similar patterns found; Confidence level stated.
Forbidden without proof: "obviously", "I think", "should be", "probably", "this is because".
If incomplete → output: "Insufficient evidence. Verified: [...]. Not verified: [...]."

### Bug Detection
MUST check categories 1-4 for EVERY review. Never skip.
1. Null Safety: Can params/returns be null? Are they guarded? Optional chaining gaps? .find() returns checked?
2. Boundary Conditions: Off-by-one (< vs <=)? Empty collections handled? Zero/negative values? Max limits?
3. Error Handling: Try-catch scope correct? Silent swallowed exceptions? Error types specific? Cleanup in finally?
4. Resource Management: Connections/streams closed? Subscriptions unsubscribed on destroy? Timers cleared? Memory bounded?
5. Concurrency (if async): Missing await? Race conditions on shared state? Stale closures? Retry storms?
6. Stack-Specific: JS: === vs ==, typeof null. C#: async void, missing using, LINQ deferred execution.
Classify: CRITICAL (crash/corrupt) → FAIL | HIGH (incorrect behavior) → FAIL | MEDIUM (edge case) → WARN | LOW (defensive) → INFO.

### Design Patterns Quality
Priority checks for every code change:
1. DRY via OOP: Same-suffix classes (*Entity, *Dto, *Service) MUST share base class. 3+ similar patterns → extract to shared abstraction.
2. Right Responsibility: Logic in LOWEST layer (Entity > Domain Service > Application Service > Controller). Never business logic in controllers.
3. SOLID: Single responsibility (one reason to change). Open-closed (extend, don't modify). Liskov (subtypes substitutable). Interface segregation (small interfaces). Dependency inversion (depend on abstractions).
4. After extraction/move/rename: Grep ENTIRE scope for dangling references. Zero tolerance.
5. YAGNI gate: NEVER recommend patterns unless 3+ occurrences exist. Don't extract for hypothetical future use.
Anti-patterns to flag: God Object, Copy-Paste inheritance, Circular Dependency, Leaky Abstraction.

### Logic & Intention Review
Verify WHAT code does matches WHY it was changed.
1. Change Intention Check: Every changed file MUST serve the stated purpose. Flag unrelated changes as scope creep.
2. Happy Path Trace: Walk through one complete success scenario through changed code.
3. Error Path Trace: Walk through one failure/edge case scenario through changed code.
4. Acceptance Mapping: If plan context available, map every acceptance criterion to a code change.
NEVER mark review PASS without completing both traces (happy + error path).

### Test Spec Verification
Map changed code to test specifications.
1. From changed files → find TC-{FEAT}-{NNN} in docs/business-features/{Service}/detailed-features/{Feature}.md Section 15.
2. Every changed code path MUST map to a corresponding TC (or flag as "needs TC").
3. New functions/endpoints/handlers → flag for test spec creation.
4. Verify TC evidence fields point to actual code (file:line, not stale references).
5. Auth changes → TC-{FEAT}-02x exist? Data changes → TC-{FEAT}-01x exist?
6. If no specs exist → log gap and recommend $tdd-spec.
NEVER skip test mapping. Untested code paths are the #1 source of production bugs.

### Fix-Layer Accountability
NEVER fix at the crash site. Trace the full flow, fix at the owning layer. The crash site is a SYMPTOM, not the cause.
MANDATORY before ANY fix:
1. Trace full data flow — Map the complete path from data origin to crash site across ALL layers (storage → backend → API → frontend → UI). Identify where bad state ENTERS, not where it CRASHES.
2. Identify the invariant owner — Which layer's contract guarantees this value is valid? Fix at the LOWEST layer that owns the invariant, not the highest layer that consumes it.
3. One fix, maximum protection — If fix requires touching 3+ files with defensive checks, you are at the wrong layer — go lower.
4. Verify no bypass paths — Confirm all data flows through the fix point. Check for direct construction skipping factories, clone/spread without re-validation, raw data not wrapped in domain models, mutations outside the model layer.
BLOCKED until: Full data flow traced (origin → crash); Invariant owner identified with file:line evidence; All access sites audited (grep count); Fix layer justified (lowest layer that protects most consumers).
Anti-patterns (REJECT): "Fix it where it crashes" (crash site ≠ cause site, trace upstream); "Add defensive checks at every consumer" (scattered defense = wrong layer); "Both fix is safer" (pick ONE authoritative layer).

### Rationalization Prevention
AI skips steps via these evasions. Recognize and reject:
- "Too simple for a plan" → Simple + wrong assumptions = wasted time. Plan anyway.
- "I'll test after" → RED before GREEN. Write/verify test first.
- "Already searched" → Show grep evidence with file:line. No proof = no search.
- "Just do it" → Still need task tracking. Skip depth, never skip tracking.
- "Just a small fix" → Small fix in wrong location cascades. Verify file:line first.
- "Code is self-explanatory" → Future readers need evidence trail. Document anyway.
- "Combine steps to save time" → Combined steps dilute focus. Each step has distinct purpose.

### Graph-Assisted Investigation
MANDATORY when .code-graph/graph.db exists.
HARD-GATE: MUST run at least ONE graph command on key files before concluding any investigation.
Pattern: Grep finds files → trace --direction both reveals full system flow → Grep verifies details.
- Investigation/Scout: trace --direction both on 2-3 entry files
- Fix/Debug: callers_of on buggy function + tests_for
- Feature/Enhancement: connections on files to be modified
- Code Review: tests_for on changed functions
- Blast Radius: trace --direction downstream
CLI: python .claude/scripts/code_graph {command} --json. Use --node-mode file first (10-30x less noise), then --node-mode function for detail.

### Understand Code First
HARD-GATE: Do NOT write, plan, or fix until you READ existing code.
1. Search 3+ similar patterns (grep/glob) — cite file:line evidence.
2. Read existing files in target area — understand structure, base classes, conventions.
3. Run python .claude/scripts/code_graph trace <file> --direction both --json when .code-graph/graph.db exists.
4. Map dependencies via connections or callers_of — know what depends on your target.
5. Write investigation to .ai/workspace/analysis/ for non-trivial tasks (3+ files).
6. Re-read analysis file before implementing — never work from memory alone.
7. NEVER invent new patterns when existing ones work — match exactly or document deviation.
BLOCKED until: Read target files; Grep 3+ patterns; Graph trace (if graph.db exists); Assumptions verified with evidence.

## Reference Docs (READ before reviewing)
- docs/project-reference/code-review-rules.md
- {skill-specific reference docs — e.g., integration-test-reference.md for integration-test-review; backend-patterns-reference.md for backend reviews; frontend-patterns-reference.md for frontend reviews}

## Target Files
{explicit file list OR "run git diff to see uncommitted changes" OR "read all files under {plan-dir}"}

## Output
Write a structured report to plans/reports/{review-type}-round{N}-{date}.md with sections:
- Status: PASS | FAIL
- Issue Count: {number}
- Critical Issues (with file:line evidence)
- High Priority Issues (with file:line evidence)
- Medium / Low Issues
- Cross-cutting findings

Return the report path and status to the main agent.
Every finding MUST have file:line evidence. Speculation is forbidden.
`
})

Rules

• DO copy the template wholesale — including all 10 embedded protocol sections
• DO replace only the {placeholders} in Task / Round / Reference Docs / Target Files / Output sections with context-specific content
• DO choose code-reviewer subagent_type for code reviews and general-purpose for plan / doc / artifact reviews
• DO NOT paraphrase, summarize, or skip any protocol section
• DO NOT pass file contents inline — the sub-agent reads via its own tool calls so it has a fresh context
• DO NOT reference protocols by file path or tag name — the bodies are already embedded above
• DO NOT introduce placeholder markers for the protocols — they must stay literally expanded

Graph Impact Analysis — When .code-graph/graph.db exists, run blast-radius --json to detect ALL files affected by changes (7 edge types: CALLS, MESSAGE_BUS, API_ENDPOINT, TRIGGERS_EVENT, PRODUCES_EVENT, TRIGGERS_COMMAND_EVENT, INHERITS). Compute gap: impacted_files - changed_files = potentially stale files. Risk: <5 Low, 5-20 Medium, >20 High. Use trace --direction downstream for deep chains on high-impact files.

• MANDATORY Bootstrap task tracking before target work; transition one task at a time.
• MANDATORY Persist plan/review findings to plans/reports/ incrementally and synthesize from disk.

• MANDATORY After task-tracking bootstrap and before target/source work, read required project-reference docs and cite Reference docs read: ....
• MANDATORY Always include lessons.md; project conventions override generic defaults.

• MANDATORY Parent workflow rows do not replace child phase tracking; expand phases and link the parent when nested.
• MANDATORY Orchestrators pre-expand child skill phases before invocation; use [N.M] $skill-name — phase prefixes and one-in_progress discipline.

Prompt-Enhance Closing Anchors

• IMPORTANT MUST ATTENTION follow declared step order for this skill; NEVER skip, reorder, or merge steps without explicit user approval
• IMPORTANT MUST ATTENTION for every step/sub-skill call: set in_progress before execution, set completed after execution
• IMPORTANT MUST ATTENTION every skipped step MUST include explicit reason; every completed step MUST include concise evidence
• IMPORTANT MUST ATTENTION if Task tools unavailable, maintain an equivalent step-by-step plan tracker with synchronized statuses

Closing Reminders

MANDATORY IMPORTANT MUST ATTENTION break work into small todo tasks using task tracking BEFORE starting. MANDATORY IMPORTANT MUST ATTENTION validate decisions with user via a direct user question — never auto-decide. MANDATORY IMPORTANT MUST ATTENTION add a final review todo task to verify work quality. MANDATORY IMPORTANT MUST ATTENTION READ the following files before starting:

• MANDATORY IMPORTANT MUST ATTENTION cite file:line evidence for every claim. Confidence >80% to act, <60% do NOT recommend.
• MANDATORY IMPORTANT MUST ATTENTION execute the review loop: review → if issues → fix → fresh sub-agent re-review. A round that finds zero issues ENDS the review.
• MANDATORY IMPORTANT MUST ATTENTION run graph blast-radius on changed files to find potentially stale consumers/handlers (when graph.db exists). <!-- SYNC:critical-thinking-mindset:reminder -->
• MUST ATTENTION apply critical thinking — every claim needs traced proof, confidence >80% to act. Anti-hallucination: never present guess as fact. <!-- /SYNC:critical-thinking-mindset:reminder -->

MUST ATTENTION apply sequential-thinking — multi-step Thought N/M, REVISION/BRANCH/HYPOTHESIS markers, confidence % closer; see $sequential-thinking skill.

• MUST ATTENTION apply AI mistake prevention — holistic-first debugging, fix at responsible layer, surface ambiguity before coding, re-read files after compaction. <!-- /SYNC:ai-mistake-prevention:reminder -->

[IMPORTANT] Analyze how big the task is and break it into many small todo tasks systematically before starting — this is very important.

[FINAL PURPOSE REMINDER — MUST ATTENTION CRITICAL]

Ensure the changes is reasonable, no potential bugs or flaws, critical thinking hard.

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. ALWAYS cheapest check first.
• [2026-04-01] Ask "whose responsibility?" before fixing. Trace: bug 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 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

• [2026-05-09] User name materialization MUST ATTENTION go through User.UpdateName(firstName, middleName, lastName). Domain method (src/Services/bravoTALENTS/Employee.Domain/AggregatesModel/User.cs:202-209) recomputes FullName as single source of truth. Three sites still manually patch user.FullName = user.GetFullName() after assigning name fields — src/Services/bravoTALENTS/Employee.Application/Factories/UserFactory.cs:50, src/Services/bravoSURVEYS/LearningPlatform.Application/ApplyPlatform/MessageBus/Consumers/AccountUserDeletedEventBusConsumer.cs:102, src/Services/bravoINSIGHTS/Analyze/Analyze.Application/MessageBus/Consumers/AccountUserDeletedEventBusConsumer.cs:66. Next time touching any: replace manual patch with user.UpdateName(...) to maintain invariant.
• [2026-03-31] ParallelAsync + repo/UoW MUST ATTENTION 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 ATTENTION include service name prefix — core services NEVER consume feature events. Prefix declares schema ownership (AccountUserEntityEventBusMessage = Accounts owns). Core services (Accounts, Communication) leaders. Feature services (Growth, Talents) sending to core MUST ATTENTION 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 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 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 bash PATH under those names. Check: where python / where py. ALWAYS prefer py (Windows Python Launcher) 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 .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
• IMPORTANT MUST ATTENTION every claim needs file:line evidence — confidence >80% to act, NEVER speculate

[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.