Pragmatic Testing

A development skill for coding agents that treats every new test as a maintenance commitment requiring concrete business or project value.

What it changes

• No automatic regression test for every bug fix or code change.
• No unsolicited tests for CSS, presentation, ordinary copy, or implementation spelling.
• Mandatory inspection of existing coverage before adding any test.
• Reuse and extend relevant scenarios or datasets before creating new files.
• Protect business rules, permissions, data integrity, integrations, and consequential shared behavior.
• Verify fixes with appropriate existing checks even when no new test is needed.
• Honor explicit requests for specific tests.

The skill instructions contain the full decision criteria and an example involving a broken datatable URL.

Install

Requires Node.js/npm for the skills CLI.

Choose agents interactively and install globally:

npx skills add arm092/pragmatic-testing -g

Install globally for every agent supported by the installer:

npx skills add arm092/pragmatic-testing -g --all

Install for selected agents:

npx skills add arm092/pragmatic-testing -g -a claude-code -a codex -a cursor

Omit -g for a project installation. Use --copy if symlinks are unavailable. Inspect the package without installing:

npx skills add arm092/pragmatic-testing --list

Compatibility

The instructions use the portable Agent Skills format, without required tools, scripts, or model-specific APIs. The installer supports Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode, Windsurf, Cline, Roo Code, and others. See its maintained supported-agent list for current coverage and installation paths.

For manual installation, copy this repository into the agent's supported skill directory under a folder named pragmatic-testing. Agents without native skill support can use the body of SKILL.md as project instructions; they do not gain automatic skill discovery from that copy alone.

Activation follows each host's skill mechanism. In Codex, explicit invocation is $pragmatic-testing, and automatic selection is enabled. Installation is not a global enforcement mechanism and does not override higher-priority instructions.

Files

• SKILL.md: portable agent instructions.
• agents/openai.yaml: Codex metadata and automatic selection policy.