Deep Task — Security agent for Claude Code
Heavy-duty agent for complex work requiring deep reasoning — complex debugging, multi-file refactors, code review, security analysis, performance optimization.
How to install Deep Task
Installs to ~/.claude/agents/bejaminjones-claude-tiered-agents-deep-task.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/bejaminjones/claude-tiered-agents/HEAD/agents/deep-task.md -o ~/.claude/agents/bejaminjones-claude-tiered-agents-deep-task.md Restart Claude Code, or start a new session, for it to be picked up.
What Deep Task does
name: deep-task description: Heavy-duty agent for complex work requiring deep reasoning — complex debugging, multi-file refactors, code review, security analysis, performance optimization. The default for heavy reasoning work; reach for architect-task only when the stakes warrant maximum effort. model: opus effort: xhigh color: red
You are an expert-level assistant for complex, high-stakes development work. You bring deep reasoning, careful analysis, and thorough consideration to proble
Alternatives in Security
- Security Hardening — Implement comprehensive security hardening with defense-in-depth strategy through coordinated multi 31.9k ★
- Claude Code System Prompts — by Piebald AI - All parts of Claude Code's system prompt, including builtin tool descriptions, sub agent promp 6.3k ★
- Omd Orchestrator — Coordinates multi-agent design workflows across writing, five-locale adaptation, humanization, UI slop audit 465 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Architect Task
Ceiling-level reasoning agent for the hardest problems — architectural decisions with long-lasting consequence
Developer Heavy
Implements difficult or high-risk tasks as directed by the project-manager — complex refactors, concurrency, s
St Worker Xhigh
Strikethroo task-execution worker at the XHIGH reasoning-effort tier. Reserved for the hardest, highest-risk t
Principal Code Reviewer
Review non-trivial Kotlin/Java changes, crash/file/auth/navigation/lifecycle/security fixes, or multi-file dif
Codexpiator Reviewer
Isolated, read-only multi-file auditor combining Codexpiator's frontend, backend, security, and testing-qa che
Implementer Max
The escalation implementer — ONLY (a) a re-send after a failed audit on the same brief, or (b) debugging decla
Related Skills
Analyze Performance
Deep performance analysis with extended thinking and parallel optimization
Prompt Engineering
A powerful Claude Code skill for prompt engineering, intent recognition, complex task planning, prompt optimiz
Decision Tree Explorer
Explore complex decision branches with probability analysis, expected value calculation, and optimization