[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.

Prerequisites: MUST READ .claude/skills/shared/understand-code-first-protocol.md AND .claude/skills/shared/evidence-based-reasoning-protocol.md before executing.

• docs/project-reference/domain-entities-reference.md — Domain entity catalog, relationships, cross-service sync (read when task involves business entities/models)

Process Discipline: MUST READ .claude/skills/shared/red-flag-stop-conditions-protocol.md — STOP after 3+ failed fix attempts or when each fix reveals a new problem.

Skill Variant: Variant of /fix — parallel multi-issue resolution using subagents.

Quick Summary

Goal: Fix multiple independent issues simultaneously using parallel fullstack-developer subagents.

Workflow:

• Triage — Classify issues and verify independence (no shared files)

• Assign — Distribute issues to parallel subagents with strict file ownership

• Execute — Subagents fix issues independently

• Merge — Review and integrate all fixes

Key Rules:

• Debug Mindset: every claim needs file:line evidence

• Issues MUST be independent (no overlapping file modifications)

• Each subagent owns specific files; no cross-boundary edits

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 parallel to fix: $ARGUMENTS

IMPORTANT: Activate needed skills. Ensure token efficiency. Sacrifice grammar for concision.

Workflow

1. Issue Analysis

• Use debugger subagent to analyze root causes

• Use /scout-ext to find related files

• Categorize issues by scope/area (frontend, backend, auth, payments, etc.)

• Identify dependencies between issues

• External Memory: Each parallel agent writes findings to .ai/workspace/analysis/{issue-name}-{agent}.analysis.md . Main agent re-reads all before coordinating fixes.

2. Parallel Fix Planning

• Trigger /plan-parallel <detailed-fix-instructions> for parallel-executable fix plan

• Wait for plan with dependency graph, execution strategy, file ownership matrix

• Group independent fixes for parallel execution

• Sequential fixes for dependent issues

3. Parallel Fix Implementation

• Read plan.md for dependency graph

• Launch multiple fullstack-developer agents in PARALLEL for independent fixes

• Example: "Fix auth + Fix payments + Fix UI" → launch 3 agents simultaneously

• Pass phase file path: {plan-dir}/phase-XX-*.md

• Include environment info

• Wait for all parallel fixes complete before dependent fixes

• Sequential fixes: launch one agent at a time

Subagent Context Discipline:

• Provide full task text — paste task content into subagent prompt; don't make subagent read plan file

• "Ask questions before starting" — subagent should surface uncertainties before implementing

• Self-review before reporting — subagent checks completeness, quality, YAGNI before returning results

4. Testing

• Use tester subagent for full test suite

• NO fake data/mocks/cheats

• Verify all issues resolved

• If fail: use debugger , fix, repeat

5. Code Review

• Two-stage review (see .claude/skills/shared/two-stage-task-review-protocol.md ):

• First: dispatch spec-compliance-reviewer to verify each fix matches its spec

• Only after spec passes: dispatch code-reviewer for quality review

• Verify fixes don't introduce regressions

• If critical issues: fix, retest

6. Project Management & Docs

• If approved: use project-manager

• docs-manager in parallel

• Update plan files, docs, roadmap

• If rejected: fix and repeat

7. Prove Fix

• MANDATORY: Run /prove-fix for EACH parallel fix

• Build code proof traces per change with confidence scores

• If any change scores < 80%, return to debug for that fix

8. Final Report

• Summary of all fixes from parallel phases

• Verification status per issue (include prove-fix confidence scores)

• Ask to commit (use git-manager if yes)

Example: Fix 1 (auth) + Fix 2 (payments) + Fix 3 (UI) → Launch 3 fullstack-developer agents → Wait → Prove each fix → Fix 4 (integration) sequential

IMPORTANT Task Planning Notes (MUST FOLLOW)

• Always plan and break work into many small todo tasks

• Always add a final review todo task to verify work quality and identify fixes/enhancements