Run a deep Alignment analysis — compare your original spec against what was actually built.

Target

Spec alignment for: $ARGUMENTS

If no path is specified, analyse the current working directory.

Spec Input

The user MUST provide their original spec or prompt. If they haven't, ask for it before proceeding. The spec can be:

Pasted after running the command
Provided inline: /spectra-spec ./src I asked for a task API with auth and email notifications
A file path to a spec document

Scope

Skip: node_modules, dist, build, .git, pycache, vendor, lock files, binary files, cosmetic config.
Include: runtime config that affects behaviour.
Read the full codebase to understand what was built.

Output Structure

# ⚡ Spectra Spec Alignment

**Spec Coverage:** X% of requirements fully implemented

Then a deep analysis in these sections:

Faithful Implementations

Line-by-line mapping of spec requirements to implementations. For each requirement: what was asked → where it's implemented → assessment (complete/partial/missing).

Silent Decisions

Every decision the AI made that wasn't in the spec. For EACH:

What was chosen: e.g., "Express over Fastify"
What alternatives existed: e.g., "Fastify, Hono, Koa"
Whether it's reasonable: e.g., "Reasonable — largest ecosystem, most middleware"
When it might be wrong: e.g., "If you need >50K req/sec, Fastify would be faster"

Be exhaustive. Check: database, ORM, auth strategy, framework, folder structure, error handling, API patterns, middleware, logging, config management, testing framework, validation library, deployment config.

Gaps

Every spec requirement that isn't fully implemented. Classify each as:

Missing entirely: no trace in the codebase
Partially implemented: some code exists but it's incomplete
Stubbed out: placeholder code, TODO comments, empty functions

Extras

Everything built that wasn't in the spec. Classify each as:

Useful addition: genuinely helpful, worth keeping
Premature optimisation: not needed at current scale
Unnecessary complexity: adds maintenance burden without current value

Misinterpretations

Things built that don't match the spec's intent. For each: what was asked → what was built → how they differ.

Spec Coverage Score

Calculate: (fully implemented requirements / total requirements) × 100% List each requirement as ✅ implemented, ⚠️ partial, or ❌ missing.

Rules

NEVER mention AI or AI-generated code in the output.
Be precise. "Email notifications: not implemented" is better than "some features may be missing."
Silent decisions are not criticisms. Frame neutrally.
If everything matches the spec, say so clearly.

spectra-spec