drewdrewthis/claude-skills

drewdrewthis/claude-skills

Skill Claude Code

Manage Boxd VM lifecycle via ssh boxd.sh — list, stop, start, fork, destroy VMs, or fork from a golden source. Use when the user says "boxd", "list my vms", "stop boxd", "pause boxd", "wake up X", "fork from golden", "destroy vm", or asks about Boxd VM status. Does NOT launch Claude sessions (use /launch-remote for…

drewdrewthis/claude-skills

Skill Claude Code

Stress-test an architecture proposal, plan, or technical idea. Invokes the devils-advocate agent to find weaknesses before you commit.

drewdrewthis/claude-skills

Skill Claude Code

Find and resume a conversation from any project. Deep-searches session content, not just titles. Gathers context and continues the work.

drewdrewthis/claude-skills

Skill Claude Code

Create a new Claude Code skill with correct structure, progressive disclosure, and token-efficient design. Use when the user wants to turn a workflow into a skill, says 'make this a skill', or asks to create/write a skill. Also triggered by /deep-reflect when it identifies a missing skill.

drewdrewthis/claude-skills

Skill Claude Code

Deep audit of Claude session history. Analyzes JSONL logs for mistakes, failed patterns, token waste, and missed opportunities. Writes findings to memory.

drewdrewthis/claude-skills

Skill Claude Code

Structured design exploration for open-ended technical questions. Combines codebase analysis, external research, strategy generation, and adversarial debate to produce a decision-ready report saved to docs/research/. Use when facing 'how should we...', 'what's the best way to...', 'deep think about', or any design…

drewdrewthis/claude-skills

Skill Claude Code

Drive a PR to mergeable state — fix CI failures and address review comments. Loops until green.

drewdrewthis/claude-skills

Skill Claude Code

Improve a skill based on how it just performed. Analyzes conversation for friction, corrections, wasted tokens, and missed steps, then patches the skill. Auto-invoke after any skill completes. Also triggered by /deep-reflect when it identifies skills that were used.

drewdrewthis/claude-skills

Skill Claude Code

Bug fix flow: readiness check → investigate → failing test (TDD) → fix → verify → review → complete. Skips planning — fixes existing behavior, not new behavior.

drewdrewthis/claude-skills

Skill Claude Code

Start implementation of a GitHub issue. Usage: /implement #123 or /implement.

drewdrewthis/claude-skills

Skill Claude Code

Systematic investigation of a production bug or performance issue. Traces symptoms to root cause through code analysis, hypothesis testing, and empirical validation.

drewdrewthis/claude-skills

Skill Claude Code

Create a worktree for a GitHub issue and launch a Claude session to implement it. One command: worktree + session.

drewdrewthis/claude-skills

Skill Claude Code

Learn from mistakes by updating project guidelines. Use when a mistake was made that should be prevented in future sessions.

drewdrewthis/claude-skills

Skill Claude Code

Log a mistake to the structured mistake log. Use when the user corrects you, when you catch your own mistake, or when /learn codifies a new rule.

drewdrewthis/claude-skills

Skill Claude Code

Audit agent config and conversation for token waste: duplication, bloat, unnecessary context loading. Produces actionable fixes.

drewdrewthis/claude-skills

Skill Claude Code

Delegate implementation work to the coder agent. Provide requirements or feature file path.

drewdrewthis/claude-skills

Skill Claude Code

Implementation and verification phase. Runs implement → sweep → review → prove-it → complete. Takes a validated feature file from /plan-work and delivers a green PR. Use when the plan is ready and you just need to build it.

drewdrewthis/claude-skills

Skill Claude Code

Metacognitive checkpoint — analyze user intent before executing. Use standalone or let orchestrate/plan invoke it. Forces critical thinking about what the user actually wants.

drewdrewthis/claude-skills

Skill Claude Code

Classify work as bug/feature/refactor, run planning, create tasks from every acceptance criterion, then hand off to implementation. The task list is the contract.

drewdrewthis/claude-skills

Skill Claude Code

Create a feature file with acceptance criteria before implementation. Use when no specs/features/.feature file exists for the work.

drewdrewthis/claude-skills

Skill Claude Code

Research and planning phase for features. Runs intent → plan → challenge → readiness gate. Produces a validated feature file ready for /deliver-work. Use before implementation when the problem needs thinking, not just coding.

drewdrewthis/claude-skills

Skill Claude Code

Prove implementation correctness by mapping every acceptance criterion to concrete evidence (passing tests, command output, code grep). Use after /review passes, before marking a PR ready. Also use standalone when you want proof that work is actually done — not just 'I think it works.'.

drewdrewthis/claude-skills

Skill Claude Code

Refactor flow: triage complexity → gather criteria → baseline → implement → verify → review → complete. Simple refactors skip challenge/prove-it. Criteria come from the issue body.

drewdrewthis/claude-skills

Skill Claude Code

Run parallel code reviews: principles-reviewer (SRP/design), hygiene-reviewer (reuse/patterns/idioms), test-reviewer (pyramid/coverage), and security-reviewer (PII/secrets). Surfaces conflicts for human decision.