cheatsheet

name	cheatsheet
description	Quick reference for all available skills and workflows. Use when you need a reminder of what's available, or when you don't know where to start.
metadata	{"type":"reference","invocation":"user-only","practice":null}

Cheatsheet

Where Do I Start?

I have an idea / feature to build:

1. Describe what you want → planning-cycles activates automatically
2. Stress-test the design → /grill-with-docs
3. Need research first? → /plan-prereqs
4. Ready to spec → spec-driven-development structures it as PLAN.md + specs
5. Build → just work (code-review, testing-guide, git-protocol all activate)
6. End session → /handoff

Something is broken: → feedback-loop-debugging activates. Build a pass/fail signal BEFORE fixing. Read references/tighten.md for loop optimization.

Starting a new session: → /read-handoff — orients you, reports what's done and next.

"What did we decide about X?": → recall search runs automatically (via steering). Searches past decisions.

Architecture feels wrong: → /architecture-deepening — finds deepening opportunities (shallow → deep modules).

Big, foggy effort (too large for one session): → /vertical-slice-planning — routes between spikes, tracer bullets, and slices based on uncertainty.

Maintenance / cleanup: → /project-cleanup (weekly) or /project-audit (drift check).

---

Commands

Command	When to use
/read-handoff	Start of session. Reads the handoff, orients you, reports what's done and what's next.
/handoff	End of session. Writes a handoff so the next session can continue without re-discovery.
/grill-with-docs	Designing something. Interrogates your plan one question at a time, updates CONTEXT.md inline.
/plan-prereqs	Before building. Identifies research, spikes, and tooling needed before implementation.
/project-cleanup	Periodic housekeeping. Promotes scratch→memory, deduplicates, processes decisions→ADR, verifies accuracy.
/project-audit	Drift check. Verifies commands work, AGENTS.md is accurate, skills are relevant.
/adopt-project	Brownfield migration. Inventories existing setup, captures special instructions, deploys.

Workflow

Start session → /read-handoff
Plan work → describe what you want (planning-cycles activates)
Stress-test → /grill-with-docs
Pre-work → /plan-prereqs
Build → just work (steering enforces hygiene + verification)
End session → /handoff
Weekly → /project-cleanup

Key Skills (activate automatically)

Skill	Triggers on
planning-cycles	"plan this", "break this down", starting a feature
feedback-loop-debugging	errors, failing tests, "help me debug", "why is this happening", root cause
code-review	"review this", "check for issues"
testing-guide	"write tests", "what should I test"
git-protocol	"commit", "push", "branch"

Project-Level Skills (install per-project)

These are NOT in global tiers. Install to a project's .kiro/skills/ when needed:

mise run add-skill -- <name>

Skill	Install when project involves...
fiction-craft	Stories, game narrative, creative prose
world-building	Fictional worlds, magic systems, game rules
presentation-writing	Slide decks, MARP, demo scripts, workshops
poc-workflow	Proofs of concept, "prove this works"
prototype-protocol	Throwaway prototypes, "let me play with it"
ux-walkthrough	Interface design, user flow evaluation
tutorial-authoring	Getting-started guides, onboarding docs
eval-criteria	Scoring rubrics, LLM-judged evals
skill-authoring	Writing or improving agent skills
enforcement-hierarchy	Deciding how to enforce agent behavior rules
project-conventions	Project-specific behavioral rules
project-winddown	Wrapping up a project, extracting lessons learned

When to suggest: during init, adopt-project, read-handoff, or when a plan includes matching work.

Extensions (auto-deploy when prerequisite met)

Extension	Commands	What it does
recall	recall search "query", recall add "fact" --type decision, recall import .memory/ --wing name	Cross-session memory — past decisions, prior work, project knowledge. Auto-primes at session start.

Prerequisites: uv tool install ./tools/recall (from a crew-research clone — the PyPI "recall" name is a squatted unrelated package). Then re-run tier deploy — recall activates automatically.

Recall quick reference

recall search "what did we decide about X"     # find past decisions
recall search "query" --type decision           # filter by type
recall add "We chose Y because Z" --type decision  # persist a decision
recall import .memory/ --wing project_name      # index project knowledge
recall import .memory/ --force --wing name      # reimport after changes
recall status                                   # see what's indexed
recall prime                                    # session-start context