Chore/bootstrap GitHub sdlc assets(DO NOT MERGE, ALREADY SPLIT)

[FScholPer]

Summary

This PR turns the repository into a thin S-CORE governance overlay that can be distributed to module repositories with Copier.

It adds:

• a reusable Copier template for applying the S-CORE governance baseline to adopter repos
• SCORE-specific governance assets and schemas
• shared coding, security, and testing instructions
• markdown hygiene checks and CI
• updated documentation for adopting and updating the overlay

It also renames the old Epic issue template to Roadmap Initiative and keeps this repository focused on lightweight SCORE-specific contracts instead of a large local agent/prompt catalog.

Why agentic SDLC

S-CORE needs a consistent way for AI-assisted development to operate across module repositories without embedding a full workflow framework in every repo.

The goal is not to centralize all agent logic here. The goal is to provide the minimum governance layer that agentic SDLC tools can rely on everywhere:

• common repository metadata
• common coding and security rules
• common artifact expectations
• a stable contract for build, test, and lint execution

This gives agentic tooling a predictable operating surface across repositories while keeping maintenance low.

How it works

This repository now acts as the S-CORE governance source of truth.

Module repositories adopt that baseline with Copier, which installs:

• shared governance instructions
• SCORE schemas
• markdown hygiene checks
• a repo-local .github/score/repo-manifest.json

That local manifest is the key integration point for agentic SDLC tooling. It tells tools how a given repository should build, test, and lint, while the distributed instructions define the guardrails those tools must follow.

The workflow framework itself can remain external (for example Spec Kit or another runtime), but it consumes this local S-CORE overlay rather than requiring each repo to maintain its own copy of governance content.

Why this shape

This keeps the architecture intentionally split:

• this repo owns S-CORE governance contracts
• adopter repos own their local execution values
• external SDLC tooling owns workflow orchestration

That makes the model easier to distribute, easier to update, and much cheaper to maintain across many repositories.

Validation

• python3 scripts/check_markdown_hygiene.py
• local Copier smoke test against a temporary adopter repo

Frank Scholter Peres frank.scholter_peres@mercedes-benz.com, Mercedes-Benz Tech Innovation GmbH
Provider Information

[lurtz]

is there somewhere defined what SDLC means?

[FScholPer]

Software Development Life Cycle

[AlexanderLanin]

PR is way to big. This would need to be one PR per file or something like that. Let's discuss in new AI interest group once we have it.

[FScholPer]

yeah but thats not practical. Initial commits of new big features are often in one pr

[FScholPer]

@AlexanderLanin how to go on here?

[FScholPer]

PR is way to big. This would need to be one PR per file or something like that. Let's discuss in new AI interest group once we have it.
@AlexanderLanin

I splitted the pr into pieces:
Priority would be:

1. PR feat: add SCORE governance core contracts #54
2. PR feat: add shared SCORE coding instructions #55
3. PR feat: add copier template for governance distribution #56
4. PR chore: add markdown hygiene checks and governance docs #57

I hope let helps

[AlexanderLanin]

@FScholPer can we move this to one of the AI repos?

[AlexanderLanin]

Thanks for the detailed write-up — the intent here is solid and the direction makes sense. A few thoughts:

The duplication concern

Almost every file in .github/instructions/, .github/references/, and scripts/ appears twice: once at the repo root and once inside template/. If template/ is what Copier distributes, the root copies seem redundant — and they create two places to maintain identical content with no mechanism keeping them in sync. Could the root just reference the template, or drop the duplicates entirely?

Generic content vs. SCORE-specific contracts

copilot-instructions.md itself says:

Avoid embedding large generic agent/prompt catalogs. Keep local content focused on SCORE-specific policy and schema contracts.

But the six instructions/*.md files (clean-code, coding-style, git-workflow, python, security, testing) are mostly standard best-practice recitations — SOLID principles, max nesting levels, commit formats. None of that is SCORE-specific, and a capable model already knows it. The genuinely valuable additions here are the schema files and the Copier setup. The generic instruction files feel like noise that'll drift out of sync and dilute the signal for tooling.

Suggestion

The thin, high-value core would be:

• repo-manifest.schema.json + agent-card.schema.json — actual contracts
• copier.yml + the Jinja templates — distribution mechanism
• Lightweight AGENTS.md/CLAUDE.md stubs in the template

Everything else could either be dropped or moved to a separate, clearly-scoped PR if it's genuinely needed.

Since this is already split and marked DO NOT MERGE, it might be worth applying this filter to the follow-up PRs before they land rather than after.

---

Generated by Claude

[AlexanderLanin]

AI-assisted review (Claude):

Two structural issues worth carrying into the split PRs.

1. Root files duplicate the template — two places to maintain.
Every file in .github/instructions/, .github/references/, and scripts/ is added both at the repo root and inside template/. copier.yml sets _subdirectory: template, so only template/ is ever distributed. The root copies exist solely for this repo's own AI tooling — but they are byte-for-byte identical to the template copies. When someone updates a guideline they'll have to update it in two places, and the markdown hygiene script won't catch drift because its default --include paths don't cover template/.

Drop the root copies and point this repo's own AI tooling at template/.github/instructions/ directly, or accept the duplication and extend the hygiene check to cover both trees.

2. copilot-instructions.md contradicts itself.
Line 22 explicitly says "Avoid embedding large generic agent/prompt catalogs." The same PR adds six instructions/*.md files (clean-code, coding-style, git-workflow, python, security, testing) that are entirely generic best-practice recitations with no SCORE-specific content. A capable model already knows SOLID principles and commit message formats; these files add token cost without adding signal.

The high-value pieces in this PR are repo-manifest.schema.json, agent-card.schema.json, copier.yml, and the Jinja templates. Consider dropping the generic instruction files or replacing them with a single thin SCORE-specific policy stub.

---

Generated by Claude

[AlexanderLanin]

AI-assisted review (Claude):

_subdirectory: template means Copier only distributes the template/ tree. Yet this PR also adds identical copies of all the same files under .github/ at the repo root. The result is two identical trees with no sync mechanism. Either remove the root copies and point this repo's AI tooling at template/.github/, or explicitly document the divergence and extend the hygiene check to cover both paths.

[AlexanderLanin]

AI-assisted review (Claude):

This line says to avoid generic catalogs, but the same PR adds instructions/clean-code, coding-style, git-workflow, python, security, and testing — all generic boilerplate with no SCORE-specific content. Either remove these files or update this rule to reflect the actual intent.