Skip to content
Back to skills

Engineering Standards

ASecurity

GENERATED ONCE PER LAYER/LANGUAGE of your chain by /adapt-to-stack, into the repo's own .claude/skills/ under a per-layer directory name (e.g. schema-standards, backend-standards, frontend-standards). The coding + review standard for <LAYER/LANGUAGE>. Loaded when writing or reviewing <LANGUAGE> code, or when running as the <layer>-specialist agent. Deliberately carries NO rules for other layers so it stays cheap to load.

  • 2 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 19, 2026
ai-agentssqltestingfrontendbackendsecurity

Security analysis

A100/100

Scanned September 19, 2026

npx -y skills add konstantinos-malavazos/claude-code-playbook --skill engineering-standards --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Engineering Standards?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Engineering Standards
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/konstantinos-malavazos-engineering-standards/badge)](https://www.skillsdirectory.com/skills/konstantinos-malavazos-engineering-standards)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: <layer>-engineering-standards
description: >-
  GENERATED ONCE PER LAYER/LANGUAGE of your chain by /adapt-to-stack, into the repo's own
  .claude/skills/ under a per-layer directory name (e.g. schema-standards,
  backend-standards, frontend-standards). The coding + review standard for
  <LAYER/LANGUAGE>. Loaded when
  writing or reviewing <LANGUAGE> code, or when running as the <layer>-specialist agent.
  Deliberately carries NO rules for other layers so it stays cheap to load.
---

# <LAYER/LANGUAGE> engineering standards

> Keep each layer's standards scoped to that layer only. A schema specialist should not
> carry frontend rules. A frontend specialist should not carry SQL index rules. Scope by
> language/layer: do not trim content, split it.

## Severity model (shared across all layers)
- `[BLOCKER]` — must fix before merge (correctness, security, data loss).
- `[MAJOR]` — should fix (design, maintainability, missing tests).
- `[MINOR]` — worth fixing (readability, small inefficiency).
- `[NIT]` — optional (style, naming).

## Correctness & design
- <e.g. layer boundaries respected; no business logic in the wrong layer>
- <e.g. inputs validated at the boundary; failure modes handled>

## <Language-specific rules>
- <e.g. async everywhere IO happens; cancellation honoured; no sync-over-async>
- <e.g. nullability handled; disposal correct>
- <e.g. for SQL: index coverage & selectivity; SARGable predicates; no N+1>

## Testing
- <what must be tested at this layer, and how (unit / integration / behaviour)>
- Test behaviour through public interfaces, not internals.

## Security & logging
- No hardcoded secrets. Mask/pseudonymize sensitive identifiers in logs.
- No logging of credentials, tokens, or raw PII.

## Compile/build correctness
- The change must build and the relevant tests must pass before hand-off.

## Vendoring note
If you have an authoritative in-house standard, vendor it here **verbatim** and pin it to
a specific commit, so the skill matches the source of truth exactly.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.