§A · SKILLS REFERENCE

Three skills, one chain.

Each skill is independently useful and they hand off cleanly, because all three share one 8-section engineering spec template. Run them in order for a brand-new effort, or jump in wherever you already have a goal or a spec. New here? Start with the step-by-step walkthrough.

`/goal-formatter`

Idea → structured, testable goal

Describe your idea in plain language and get back a structured, testable goal: a measurable end state, success criteria, a verification method, constraints, and a test-coverage shape. It self-reviews the draft against a quality rubric and revises before it shows you anything, so the goal you receive has already been checked for testability.

Add a spec-style ask ("and a spec", "spec this out") and it also emits a draft 8-section spec, marking anything the description does not determine as [GAP] for goal-spec to resolve.

When to use it

"Write me a goal for X" or "create a goal for X".
You want to frame loose work as a measurable objective before building.
You want a draft spec to hand straight to goal-spec.

Invoke

/goal-formatter in Claude Code, or the equivalent for your agent (see the usage page).

`/goal-spec`

Goal → complete engineering spec

Takes the goal (or a goal plus draft spec) and produces a complete engineering specification across all eight sections. Its defining behaviour: it asks you gap-filling clarifying questions grounded in what you said you wanted, so each question reads as "to satisfy your goal of X, which of these?" rather than a generic survey.

It resolves every gap by asking, or by recording a sensible-default assumption, then self-reviews the spec against a quality rubric: success criteria must be numeric with a named test method, every requirement testable, no unstated assumptions.

When to use it

"Spec this goal", "flesh out the spec", "define the specification".
You have a goal that needs to become an implementable design.
You have a draft spec with gaps that need resolving.

Pairs with

goal-formatter (upstream) and spec-review (downstream).

`/spec-review`

Spec → hardened spec (autonomous)

Hardens the spec autonomously. Each round spawns a roundtable of independent engineering-discipline reviewers (Architect, Security, Test/QA, Operations, Product) as separate subagents that critique the spec from their own expertise. The orchestrator triages findings, revises the spec in place, and loops until no CRITICAL or MAJOR findings remain (hard cap: 6 rounds). There are no user checkpoints: you get the converged spec and a round-by-round summary at the end.

The personas are domain-adaptive: Security and Operations auto-remap for physical products, data/ML, or process specs, so you do not hand-map them in the invocation. If your client has no subagents, it can run the personas inline (one model voicing each), faster but less independent.

When to use it

"Review this spec", "harden this spec", "run the roundtable on this spec".
You want a spec battle-tested before implementation.

Exit artifact

A ready-to-run /goal execution prompt that hands the goal plus hardened spec to an executor of your choice. Self-contained, with no dependency on any other skill.

INSTALL THE KIT →