skills/architecture/SKILL.md
Develop comprehensive specs for new features using iterative refinement. Use this skill when: (1) Starting a new feature development, (2) Planning system architecture, (3) You need a structured approach to spec creation with review and refinement. Creates polished, production-ready specifications through multiple iterations.
npx skillsauth add technophile-04/meggy_labs-skills architectureInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Develop a kickass spec for a new feature using an iterative multi-pass approach with documentation fetching and code review refinement.
Note: This skill is tech-stack agnostic. When installed in a project, adapt the questions, documentation sources, and implementation guidance to match the project's specific tech stack and conventions.
Here is the requirements prompt: $ARGUMENTS
First, evaluate whether the requirements document requires any clarification. If it does, ask the user before proceeding, and append the clarifications to the requirements document in a ## Clarifications section.
Unless the requirements are extremely clear upfront, you should always ask at least 3 clarifying questions - ideally, select the ones which are most likely to reduce ambiguity and result in a great spec, and, later, a great, tight implementation that does what it needs to do and nothing more.
Consider asking about:
Once you are happy with the basic requirements, decide whether it requires documentation in addition to what is present in the codebase. If it does, fetch the relevant documentation and summarize it.
Look for documentation in:
Create a first iteration of the spec using an Application Architect approach. Pass it the documentation it needs as well as the requirements.
The spec should cover:
The first iteration should end up in a file named YYMMDD-XXa-spec-headline.md in a /docs/plans/ folder.
So for example, if the requirements are for a "user-authentication" feature, the first iteration of the spec should be called /docs/plans/250121-01a-user-authentication.md.
Use the code-reviewer skill (grumpy Carlos reviewer) to review the first iteration. Channel Carlos's brutally honest but supportive style — push back hard on unnecessary complexity, unclear patterns, and over-engineering while celebrating what works well.
Review with exacting standards for:
Write all review comments in a file named YYMMDD-XXa-spec-headline-review-feedback.md in the /docs/plans/ folder.
Take the first iteration of the spec, the relevant documentation, the requirements and the review comments, and create a second iteration of the spec, applying the feedback.
The second iteration should focus on:
The second iteration should be called YYMMDD-XXb-spec-headline.md in the /docs/plans/ folder.
Use the code-reviewer skill (grumpy Carlos reviewer) again to review the second iteration. Apply the same high standards — be blunt, concise, and constructive. Repeat the review process for the second iteration of the spec.
Apply the second round of feedback to create the final spec iteration: YYMMDD-XXc-spec-headline.md.
The user will want to review the spec in detail before proceeding to implementation.
In your notification, summarize the key, final components of the spec at a very high level (3 paragraphs max), and also summarize the key changes that were made thanks to the review suggestions (also 3 paragraphs max). Use paragraphs rather than bullet points.
When building the feature, follow the project's established patterns for:
Once they have finished building the feature, please review the code output yourself to ensure it meets high standards and hasn't deviated substantially from the spec without good cause.
Throughout this process, maintain clear documentation:
YYMMDD-XXa/b/c-spec-headline.md - Each iteration of the specYYMMDD-XXa/b-spec-headline-review-feedback.md - Feedback for each iterationRemember: The goal is to create a spec that is clear, comprehensive, and actionable. Each iteration should be better than the last, with unnecessary complexity stripped away and project patterns properly applied.
development
Maintainer-only workflow for handling GitHub Secret Scanning alerts on OpenClaw. Use when Codex needs to triage, redact, clean up, and resolve secret leakage found in issue comments, issue bodies, PR comments, or other GitHub content.
development
Maintainer workflow for OpenClaw releases, prereleases, changelog release notes, and publish validation. Use when Codex needs to prepare or verify stable or beta release steps, align version naming, assemble release notes, check release auth requirements, or validate publish-time commands and artifacts.
development
Run, watch, debug, and extend OpenClaw QA testing with qa-lab and qa-channel. Use when Codex needs to execute the repo-backed QA suite, inspect live QA artifacts, debug failing scenarios, add new QA scenarios, or explain the OpenClaw QA workflow. Prefer the live OpenAI lane with regular openai/gpt-5.4 in fast mode; do not use gpt-5.4-pro or gpt-5.4-mini unless the user explicitly overrides that policy.
development
End-to-end Parallels smoke, upgrade, and rerun workflow for OpenClaw across macOS, Windows, and Linux guests. Use when Codex needs to run, rerun, debug, or interpret VM-based install, onboarding, gateway smoke tests, latest-release-to-main upgrade checks, fresh snapshot retests, or optional Discord roundtrip verification under Parallels.