mac

Use when deciding how deeply to review a PR, which reviewers to involve, or whether to accept a large agentic PR

Plan test coverage and the TDD workflow for a GitHub issue through a grill-me interview, then produce an execution-ready plan doc (grounding, test locations, red/green steps, PR breakdown, review gates, task-list bookkeeping) at the repo root. Use when asked to plan a TDD implementation, plan which tests to add for an issue, or scope out test coverage before writing code.

Use ONLY when Nico explicitly invokes it (typically via /pdp-confidence-update) to write, prepare, or reconstruct the weekly PDP confidence update for a project channel. Do NOT trigger automatically from general Slack/status questions. Reconstructs the channel since the last `!pdp progress` post and drafts a ready-to-post update in Nico's house style.

Produces Nico's daily stand-up reply — "Yesterday", "Today", and "Blockers or risks" sections in his house style, organized by his current P1/P2/P3 epics with everything else under "Also" — then posts each as a separate Geekbot reply in his Slack DM once he agrees on the wording. Use when Nico asks to write his daily stand-up, stand-up message, Geekbot reply, or to summarize yesterday's work for stand-up.

Use when reviewing a PR in camunda/camunda that touches AGENTS.md, skills under .claude/skills/, the way the project uses ADRs, or the way the project uses architecture docs (not new ADRs or architecture docs themselves) — i.e., Agent-Ready initiative PRs

Use when user shares an error message, stack trace, or log snippet to investigate — or when investigating an incident that has a concrete error artifact (log line, exception, failed assertion). Skip when the incident has no error artifact, e.g. performance regressions, throughput/latency drops, capacity issues, or behavioral discrepancies without logs.

Refines an existing skill for effectiveness (does it achieve its goal?), efficiency (tokens spent doing so), and discoverability (does Claude trigger it when relevant?). Use when the user wants to optimize, refactor, audit, or tighten a skill, or invokes optimize-skill.

Produce a deterministic local reproducer for a bug described by the user. Use when the user shares an incident, bug report, or behavior description and wants to reproduce it locally. Output is a reproducer (test or script) for later root-cause analysis — RCA itself is out of scope.

Conducts a comprehensive research across the codebase to answer user questions by spawning parallel sub-agents and synthesizing their findings.

Creates detailed implementation plans through an interactive, iterative process.

Implements an approved technical plan from `thoughts/plans/`. These plans contain phases with specific changes and success criteria.

Review a GitHub epic issue to surface unclarity, ambiguousness, or vagueness in its User Problem and User Stories sections. Use when user wants to review an epic, check an epic for gaps, or mentions "review epic".

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

Validates that an implementation plan was correctly executed, verifying all success criteria and identifying any deviations or issues.