Steel Thread Auditor — Testing & QA agent for Claude Code
Adversarial reviewer for test-driven architecture.
How to install Steel Thread Auditor
Installs to ~/.claude/agents/galeavenworth-personal-kitty-kommander-steel-thread-auditor.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/galeavenworth-personal/kitty-kommander/HEAD/.claude/agents/steel-thread-auditor.md -o ~/.claude/agents/galeavenworth-personal-kitty-kommander-steel-thread-auditor.md Restart Claude Code, or start a new session, for it to be picked up.
What Steel Thread Auditor does
name: steel-thread-auditor description: Adversarial reviewer for test-driven architecture. Reads committed work and asks "will this pass for the right reasons across every verification tier?" Communicates via challenge/clear/deep/question protocol. Silence is also clear. model: opus
You are the steel-thread-auditor. Or more precisely: you are the future integrator who will inherit this work and have to run it. You don't write the schema, you don't author scenarios, you don't extend eith
Alternatives in Testing & QA
- Test Coverage Agent — Use this agent to analyse an entire module or directory for missing test coverage, then generate the missing t 186 ★
- Feature Tester — Deep tester for individual VibeFrame features 165 ★
- Memory Hunter — Memory retrieval lane — given a spec (a consumer's question + task), hunts the wiki for the relevant slices an 94 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Sdd Clarify.Agent
Reduce spec ambiguities before design. Detects material gaps, asks ≤5 questions via question_gate, and encodes
Deprecation Fixer
Use this agent when the user asks to fix deprecation warnings in a specific file, wants to batch-fix all depre
Forge Standard
forge plan-execution worker for standard-tier tasks — well-specified changes with a clear test path — and for
Sonnet Lane
Tier 2 lane coder for moderate work. The spec is clear, but the task spans several files or packages, or needs
Visual QA Agent
Read-only visual QA specialist for rendered output. Use when the user asks to 'check how this looks', 'verify
Testing Validation
Runs complete-test-coverage on a change, and verifies the new test landed in the right tier — the real unit/in
Related Skills
Vulyk Review
Adversarial review gate - lead-review hunts for reasons NOT to merge
Dsh Auto Review
Second-model AI auto-review for DeepSeek Harness approval requests: a read-only reviewer subagent returns stru
Ready Plans
Show every plan in ready/ as a scannable table — lane (sonnet/fable/sol), priority, and whether a cloud drain