Using dev-skills

Entry map for the local dev-skills toolkit. It does not prescribe a methodology; it groups skills by the input they apply to so the right one can be invoked.

Session-start default. Consulting this map is the default first action of a session, not an optional aid. If a task begins without it, invoke it before any repo exploration or code change.

Ladders

Each ladder is a cluster of skills that apply to the same artifact or phase. Use the input you have to pick the ladder; invoke the matching skill via the Skill tool.

TDD ladder

For test-first work. Applies on a list of test names to evaluate, a failing test to drive, or a request for RED/GREEN discipline.

Spec ladder

For going from a vague feature ask to a reviewable contract. Applies on a feature request, a component boundary to define, or a service spec needing explicit collaborator boundaries.

Architecture review ladder

For reviewing design or implementation against architectural intent. Applies on a design doc PR, an architecture spec, a sketch of system flow, an implemented component up for critique, or a REST contract change.

Debugging ladder

For investigating broken behavior. Applies on a bug report, a stack trace, or the need to trace execution before changing code.

Planning ladder

For aligning before code is written. Applies on a new task with unclear scope, an exploration request, a review finding that needs an implementation plan, an implementation plan being authored or edited, or a gate before commit/push. Use implementation-plan for plan authoring/declaration checks and process-gates for execution gates.

Release / CI ladder

For shipping work. Applies on ADRs, per-unit releases, PR CI monitoring, or docker compose / deploy changes.

Swift / iOS ladder

For Swift packages and iOS. Applies on composition root wiring, protocol ownership, Package.swift, time-dependent tests, xcodebuild runs, or simulator install / settings tasks.

Skill maintenance

For the toolkit itself. Applies on the first message of a session (replay-before-execute) or on evaluating an existing skill.

Ladder selection cues

You have…	Ladder
A bug report or stack trace, analysis wanted	Debugging
A bug report with a fix/deploy requested	Treat as delivery: investigate inline, no analysis-report skill
A vague feature request	Spec
An approved review finding	Planning
A test name list or RED/GREEN ask	TDD
A PR open and CI running	Release / CI
A design doc or arch sketch in review	Architecture review
A Swift package or iOS simulator task	Swift / iOS
None of the above	Ask one clarifying question, then re-check

What this skill is not

• Not a methodology. Ladders compose per task; there is no required sequence.
• Not a catalog. It points to ladders; the invoked skill governs.
• Not exhaustive. Skills outside any ladder are discoverable via their own descriptions; new skills land outside any ladder before they cluster.