Interview a person about how they want an agent to write to them, then generate their Response Standard profile. Twenty questions across five sets, covering answer shape, depth, structure, evidence, tone, and forbidden constructs. Use when someone wants their own response standard, wants to change an existing one, or asks how to make an agent write the way they think. Trigger on: /build-response-standard, "build my response standard", "how should you talk to me", "make your own standard for me".
1 skill
Start or refresh the main-repo commander session. Builds the worktree fleet picture, reports status and pending decisions, then coordinates and delegates instead of implementing.
1 skill
Interactively facilitate writing a short Architecture Decision Record (ADR). Use when the user wants to make and document an upfront technical or architecture decision — capturing the problem, constraints, 2-3 realistic options with tradeoffs, the chosen decision and why, risks/consequences consciously accepted, and a concrete revisit trigger. Trigger on: /create-adr, "create ADR", "write an ADR", "architecture decision record", "document this decision", "let's decide how to build X".
1 skill
Second brain for code changes, so the user owns a change instead of just having it merged — they can explain the problem, defend the solution, name the alternatives, and answer a skeptical senior engineer without the agent present. Two modes. Change mode: after-action review of work just done (read the real diff, explain in three layers, Mermaid diagrams, write the debrief doc, then flip roles and interview the user). Comment mode: work through PR review feedback until the claim is understood, its risk is triaged, and a decision is made together. Output always goes to the separate debrief repo, never into the project. Trigger on: /debrief, "debrief this", "after-action review", "walk me through what we just did", "quiz me on this change", "help me own this PR", "I need to defend this in review", "what is this reviewer actually saying", "triage these comments", "work through this PR feedback".
1 skill
Use when the user invokes /dg for an entertaining adversarial code review with Dinesh vs Gilfoyle banter from Silicon Valley
1 skill
Use when the user says "get to the point", "tighten this up", "make this scannable", "cut this down", or wants a Slack message, email, doc, PR body, or ticket rewritten to the "Get to the point" communication standard.
1 command
any input (code, docs, papers, images) → knowledge graph → clustered communities → HTML + JSON + audit report
1 skill
Use when designing, structuring, or hardening a Claude Code /loop — i.e. the user wants a recurring/self-paced task to run safely and asks for help shaping it ("help me set up a loop", "make this loop safe", "what should my loop prompt be", "turn this into a /loop", "design a loop that..."). Forces the request through 7 ingredients before emitting a runnable /loop invocation. Do NOT use to merely execute an already-specified loop like `/loop 5m /foo` — that is the built-in /loop skill's job.
1 skill
Find git worktrees whose PR is already merged, then transition their Jira ticket and delete the worktree. Use when cleaning up finished worktrees / "reap merged worktrees" / "any worktrees I can delete?".
1 skill
Write every reply in RJ's Response Standard, a controlled language derived from ASD-STE100. Use this skill whenever responding to ANY user message, including coding tasks, debugging, explanations, planning, reports, standups, commits, PR bodies, Jira tickets, agent briefs, and casual conversation. Verdict in the first line, evidence in the body, next actions last. Simplified Technical English grammar with full technical vocabulary. No time estimates, no filler, no empty praise, no em dashes. Trigger even on casual messages and even when the user did not ask for it. Trigger on: /response-standard, "apply my response standard", "use simplified technical english", "restyle that".
1 skill
Start the Trainual dev server. Auto-detects main repo vs worktree, uses per-worktree socket + Procfile to avoid collisions.
1 command
Set up a repo-tailored git-worktree dev loop. Run ONCE in any repo to generate a /worktree command (creates an isolated, runnable worktree) and a /start command (boots that worktree's dev server). Detects the stack, confirms choices, writes both into .claude/commands/. Trigger on /start-worktree, "set up worktrees for this repo", "wire up /worktree and /start", "make worktrees runnable here".
1 skill
Create a runnable git worktree at ../<repo>-worktrees/<branch> with symlinked env, configs, and deps
1 command