fix-hard

[IMPORTANT] Use TaskCreate 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 ask user whether to skip.

Understand Code First — Search codebase for 3+ similar implementations BEFORE writing any code. Read existing files, validate assumptions with grep evidence, map dependencies via graph trace. Never invent new patterns when existing ones work. MUST READ .claude/skills/shared/understand-code-first-protocol.md for full protocol and checklists.

Evidence-Based Reasoning — Speculation is FORBIDDEN. Every claim needs file:line proof. Confidence: >95% recommend freely, 80-94% with caveats, <80% DO NOT recommend — gather more evidence. Cross-service validation required for architectural changes. MUST READ .claude/skills/shared/evidence-based-reasoning-protocol.md for full protocol and checklists.

• docs/project-reference/domain-entities-reference.md — Domain entity catalog, relationships, cross-service sync (read when task involves business entities/models) (content auto-injected by hook — check for [Injected: ...] header before reading)

Estimation Framework — SP scale: 1(trivial) → 2(small) → 3(medium) → 5(large) → 8(very large, high risk) → 13(epic, SHOULD split) → 21(MUST split). MUST provide story_points and complexity estimate after investigation. MUST READ .claude/skills/shared/estimation-framework.md for full protocol and checklists.

Skill Variant: Variant of /fix — deep investigation with subagents for complex issues.

Quick Summary

Goal: Systematically diagnose and fix complex bugs using parallel subagent investigation.

Workflow:

1. Scout — Use scout/researcher subagents to explore the issue in parallel
2. Diagnose — Trace root cause through code paths with evidence
3. Plan — Create fix plan with impact analysis
4. Fix — Implement and verify the fix

Key Rules:

• Debug Mindset: every claim needs file:line evidence
• Use subagents for parallel investigation of multiple hypotheses
• Always create a plan before implementing complex fixes

[MANDATORY] Read .claude/skills/shared/root-cause-debugging-protocol.md BEFORE proposing any fix. Responsibility attribution and data lifecycle tracing are required.

Debug Mindset (NON-NEGOTIABLE)

Be skeptical. Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence percentages (Idea should be more than 80%).

• Do NOT assume the first hypothesis is correct — verify with actual code traces
• Every root cause claim must include file:line evidence
• If you cannot prove a root cause with a code trace, state "hypothesis, not confirmed"
• Question assumptions: "Is this really the cause?" → trace the actual execution path
• Challenge completeness: "Are there other contributing factors?" → check related code paths
• No "should fix it" without proof — verify the fix addresses the traced root cause

⚠️ MANDATORY: Confidence & Evidence Gate

MANDATORY IMPORTANT MUST declare Confidence: X% with evidence list + file:line proof for EVERY claim. 95%+ recommend freely | 80-94% with caveats | 60-79% list unknowns | <60% STOP — gather more evidence.

Ultrathink to plan & start fixing these issues follow the Orchestration Protocol, Core Responsibilities, Subagents Team and Development Rules: $ARGUMENTS

Workflow:

If the user provides a screenshots or videos, use ai-multimodal skill to describe as detailed as possible the issue, make sure developers can predict the root causes easily based on the description.

Fullfill the request

Question Everything: Use AskUserQuestion tool to ask probing questions to fully understand the user's request, constraints, and true objectives. Don't assume - clarify until you're 100% certain.

• If you have any questions, use AskUserQuestion tool to ask the user to clarify them.
• Ask 1 question at a time, wait for the user to answer before moving to the next question.
• If you don't have any questions, start the next step.

⚠️ Validate Before Fix (NON-NEGOTIABLE): After root cause + plan creation, MUST present findings + proposed fix plan to user via AskUserQuestion and get explicit approval BEFORE any code changes. No silent fixes.

Fix the issue

Use sequential-thinking skill to break complex problems into sequential thought steps. Use problem-solving skills to tackle the issues. Analyze the skills catalog and activate other skills that are needed for the task during the process.

1. Use debugger subagent to find the root cause of the issues and report back to main agent. 1.5. Write investigation results to .ai/workspace/analysis/{issue-name}.analysis.md. Re-read ENTIRE file before planning fix.
2. Use researcher subagent to research quickly about the root causes on the internet (if needed) and report back to main agent.
3. Use planner subagent to create an implementation plan based on the reports, then report back to main agent.
4. 🛑 Present root cause + fix plan → AskUserQuestion → wait for user approval.
5. Then use /code SlashCommand to implement the plan step by step.
6. Final Report:

• Report back to user with a summary of the changes and explain everything briefly, guide user to get started and suggest the next steps.
• Ask the user if they want to commit and push to git repository, if yes, use git-manager subagent to commit and push to git repository.

• IMPORTANT: Sacrifice grammar for the sake of concision when writing reports.
• IMPORTANT: In reports, list any unresolved questions at the end, if any.

REMEMBER:

• You can always generate images with ai-multimodal skills on the fly for visual assets.

• You always read and analyze the generated assets with ai-multimodal skills to verify they meet requirements.

• For image editing (removing background, adjusting, cropping), use media-processing skill as needed.

• After fixing, MUST run /prove-fix — build code proof traces per change with confidence scores. Never skip.

Closing Reminders

• MUST break work into small todo tasks using TaskCreate BEFORE starting
• MUST search codebase for 3+ similar patterns before creating new code
• MUST cite file:line evidence for every claim (confidence >80% to act)
• MUST add a final review todo task to verify work quality
• MUST STOP after 3 failed fix attempts — report outcomes, ask user before #4 MANDATORY IMPORTANT MUST READ the following files before starting:
• MUST READ .claude/skills/shared/understand-code-first-protocol.md before starting
• MUST READ .claude/skills/shared/evidence-based-reasoning-protocol.md before starting
• MUST READ .claude/skills/shared/estimation-framework.md before starting