Skip to content
Back to skills

Hld Review

ASecurity

Review a high-level design (HLD): a design doc, RFC, architecture proposal or "how should we build X" plan. Use before implementation starts, or when asked to review, challenge or sign off a system design.

  • 5 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 11, 2026
ai-agentsrustgosecurity

Security analysis

A100/100

Scanned November 11, 2026

npx -y skills add thameema/memnos --skill hld-review --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Hld Review?

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

Security grade badge for Hld Review
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/thameema-hld-review/badge)](https://www.skillsdirectory.com/skills/thameema-hld-review)

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: hld-review
description: Review a high-level design (HLD): a design doc, RFC, architecture proposal or "how should we build X" plan. Use before implementation starts, or when asked to review, challenge or sign off a system design.
---

# HLD review

Review the design against the real system, not against the doc's own claims. Read the design first, then read the code and config it says it touches. A finding that cites a file or a number beats one that cites a principle.

## Steps

1. **State the goal and non-goals** in two lines. If the doc doesn't make them clear, that is finding #1.
2. **Check the boundaries.** Which components own which data? Where does each request cross a process, network or trust boundary? Name every new boundary and what authenticates it.
3. **Walk the failure modes.** For each dependency: what happens when it is slow, down, or returns garbage? Look for retries without backoff or idempotency, missing timeouts, unbounded queues, and partial failure across two writes.
4. **Check data and migration.** Schema changes, backfills, rollback path, and whether old and new versions can run side by side during a rolling deploy.
5. **Check scale and cost** with numbers: expected load, the hot path, the largest table or fan-out, and what grows without bound.
6. **Check security and privacy**: who can read or write what, where secrets live, what gets logged, tenant or namespace isolation.
7. **Check it fits the codebase**: does something equivalent already exist? Does it follow the existing layering and conventions, or does it quietly introduce a second way of doing the same thing?
8. **Check operability**: how will anyone know it works in production (metrics, alerts, logs), how is it rolled out and rolled back.

## Output

- **Verdict:** approve / approve with changes / rework.
- **Blocking findings**, ranked, each with: the claim in the design, the evidence (file, line or number), and the concrete change that fixes it.
- **Non-blocking suggestions**, short.
- **Open questions** the author must answer.
- **What you did not check** (e.g. no load numbers were available, the dependency's SLA was not read).

Do not rewrite the design. Do not pad with praise. If the design is sound, say so in one line.

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.

View all in ai-agents →

Comments

Loading comments…

Newsletter

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