pilot v0.10.0 field guide · sdlc ← overview

USE IT
RIGHT.

Pilot rides the whole lifecycle — you just describe the work in your own words. This page shows what to say at each SDLC phase, which skill pilot routes to, which gates protect you, and what you get out.

rule 1
Never memorize skill names. Describing intent ("build…", "it's broken…", "ship it") routes correctly. Naming a skill literally (tdd, diagnose, graphify) is an override, not a requirement.
rule 2
Gates want evidence, not promises. A "done" claim needs a real captured test run; >20-line edits need a plan artifact first. Work with them — they're what makes the output trustworthy.
rule 3
Stuck or blocked wrongly? pilot off skips the next gate fire, /pilot-status shows what's armed, /pilot-doctor diagnoses the install.

pick your mode

Three ways to drive. They compose — start hands-off, drop to manual when a phase needs your judgment.

default · conversational

Describe the work

"add rate limiting to the API" → pilot detects the phase from your words + project state and invokes the right skill. You approve plans, review diffs, say "go" between phases.

explicit · power user

Name your tools

"use context7 for the docs, plan with writing-plans, tdd it, verify with playwright" → every named tool becomes a phase in a chain, executed in order. No keyword scoring.

hands-off · one requirement

Autopilot

/pilot-autopilot "<requirement>" → the whole loop runs unattended and stops exactly twice: plan approval and ship approval. Reply approved — your prompt is machine-witnessed.

the lifecycle, case by case

Each card: what you type, where pilot routes, which gates are live, what artifact you walk away with.

01

Kick off a project

new project — a React Native app for split expenses init this repo
routes
initgsd-new-project (structured, multi-session) — fires automatically when there's no CLAUDE.md
gates
floor scaffolded integrity-check
you get
CLAUDE.md, .planning/ project state, CI + pre-commit quality-floor templates ready to activate with /pilot-floor
02

Discover & frame requirements

I have an idea — daily briefing from my calendar. worth building? build a notifications feature for missed workouts
routes
viability → office-hours (product interrogation); code intent → grill-with-docsto-prd. Ambiguities become one focused question, not three assumptions
gates
clarify scan
you get
a design doc / PRD with success criteria — before any code exists
03

Plan with acceptance criteria

plan this challenge this plan — scope up or down?
routes
superpowers:writing-plans (single-session) or gsd-plan-phase (multi-day); ceo-review to stress the plan strategically
gates
G1 plan-gate AC-first
you get
a phased plan + .pilot/acceptance.md (AC-001… checkboxes) — the ledger verify-gate later refuses to close unchecked
04

Build — logic and UI

implement slice 1 build the settings screen UI
routes
logic → tdd (red-green-refactor, one commit per tracer slice); UI → ui-ux-pro-max / frontend-design; unfamiliar library → context7 pulls current docs first
gates
G1 plan-gate G3 commits G7 no any G8 no console.log G15 safety
you get
small conventional commits, each slice tested as it lands
05

Test & verify

done — ready to commit actually test it in the browser
routes
verify runs the suite so the result is captured; UI changes get driven for real — playwright-cli first (open → snapshot → click → eval → screenshot), playwright MCP fallback
gates
G14 verify-gate blocks AC checkoff
you get
"done" only clears with a real captured pass + every AC checked. Prose never counts — the gate blocked its own author twice
06

Debug & performance

login throws a 500 after the deploy the list screen is slow — profile it
routes
diagnose — reproduce → hypothesize → instrument → fix → regression-test. Hypothesis before edit is non-skippable; no sleep-based band-aids
gates
G11 hypothesis-first G12 no sleep in tests
you get
a root-cause fix with a regression test, not a patched symptom
07

Review & security

review this branch before I open the PR security review — this touches auth
routes
superpowers:requesting-code-review (independent read); auth/crypto/network diffs trigger security-review — mandatory before Ship; github MCP reads real PR/CI state
gates
review before ship 6.5 security
you get
findings addressed pre-merge; sensitive paths audited with SAST + secret scan backing
08

Ship, release, deploy

ship it cut a release and deploy to prod
routes
gsd-ship (PR → review → merge); release artifacts → version-bump (semver + changelog + tag); prod-bound → pre-deploy-checklist fires first, post-deploy-monitor watches after
gates
verify + review green 7.75 pre-deploy
you get
PRE-DEPLOY.md (secrets, env, rollback path), a tagged release, POST-DEPLOY.md watching error rate + latency
09

Operate & maintain

upgrade react-navigation — major version this module is messy and hard to change graphify the repo
routes
deps/schema → migration-safety (MIGRATION-SAFETY.md before pre-deploy); smells → improve-codebase-architecture; graphify maps the codebase into a queryable knowledge graph — and re-maps each cycle's delta at Capture
gates
7.5 migration SCA audit
you get
safe upgrades, scoped refactors, and a knowledge graph that answers "what breaks if I change X?"
10

Or skip the conducting: autopilot

/pilot-autopilot "users can export their expense history as CSV"
runs
frame → plan → ⏸ your plan approval → build (TDD) → verify → fix loop (max 3 rounds, then halts with a report) → review → ⏸ your ship approval → ship → capture
gates
G16 cycle gate witnessed approvals all of the above still apply
notes
State survives crashes — say continue to resume. Branch-scoped, so teammates run parallel cycles. autopilot off aborts. Approvals only count from your actual prompt (or /pilot-approve) — the model can't approve itself

working as a team

The habits that make pilot work across more than one keyboard.

Day one

bash dev/bootstrap-team.sh --check installs/verifies the pinned skill constellation, then /pilot-doctor. Ramp with docs/team-onboarding.md — ten minutes.

Per-repo profile

.pilot.json sets style + strictness + your test runner (test_patterns) so verify-gate recognizes real runs. Nobody inherits another dev's communication register.

Shared scoreboard

Turn on team.shared_outcomes; dev/outcome-report.sh shows the team's first-pass-verified rate per user. High block counts = "done" is being claimed without tests.

Prove it works

Run the A/B protocol in docs/ab-method.md — pilot-on vs pilot-off weeks, binary scoring. Keep gates that earn their keep; cut the ones that don't.