Skip to content

v2.27.0 ·

Automotive suppliers already have the trail ASPICE asks for

If you write software for a carmaker, you get assessed against Automotive SPICE on almost every contract. The assessor asks for traceability both ways: requirements to architecture to code to tests, kept consistent.

AIDLC already keeps that chain for every change. Numbered criteria in requirements.md, the design in design.md, task links in tasks.md, and a testing gate that checks each criterion has a test. Nothing presented it in the terms an assessor reads, so a supplier would rebuild it by hand before the assessment.

What changed

There is now a third compliance pack, aspice, beside SOC 2 and ISO 27001. It targets Automotive SPICE 4.0 and covers the six software engineering processes, SWE.1 to SWE.6.

aidlc compliance add aspice

Sixteen outcomes the trail can show are controls, each pointing at the file that shows it. Requirements outcomes point at requirements.md. Architecture and detailed design point at design.md or tasks.md. Verification points at test-results.md. A control id like SWE1.3 means outcome 3 of SWE.1.

Everything else is listed as excluded, with a reason: 52 entries. That includes the 17 other software outcomes, with a note on what the trail cannot show, and every management, supply and support process.

You will also notice a new question in aidlc init. This is the first pack scoped to an industry, so the wizard now asks which industry the project serves, and offers the pack only under automotive.

Why this way

ASPICE rates how an organisation works across its projects. A trail per change is good input to that rating. It is not the rating. So the report says, every time, that the pack gives process evidence per change and rates no capability level. A test fails if any sentence in the pack or the docs calls a project ASPICE compliant or says it reaches a capability level.

I also kept the model's text out. The pack paraphrases each outcome in its own words and cites process and outcome numbers only. The model itself comes from the VDA.

No CLI code changed. Packs are found by directory, so the pack is content and tests.

What it does not do

No compliance practitioner has reviewed it. The outcome list is my reading of the 4.0 model, not checked against the VDA text, and every report carries that notice.

It does not catch an artifact edited after its phase closed without aidlc amend, so it cannot prove the requirements, the design and the shipped code still agree. Functional safety (ISO 26262) and cybersecurity (ISO/SAE 21434) are out.

The change cost 21.9M tokens, about $8.33 at API price.