Enforce a three-layer verification pipeline before writing implementation code.
npx skills add https://github.com/anhtrinh919/claude-code-sdd --skill code-harness
Please help me install this Agent Skill. Skill: code-harness Source: https://github.com/anhtrinh919/claude-code-sdd/tree/main/skills/code-harness Command: npx skills add https://github.com/anhtrinh919/claude-code-sdd --skill code-harness
A three-layer discipline pipeline (Spec-Light + TDD-Guard + Verify-Script) that prevents the "AI claims done, doesn't work, non-dev can't verify, debug spiral" failure mode. Trigger this skill ALWAYS before writing any implementation code for a non-developer operator: whenever the user says "let's build/implement/add/fix/change", invokes /dev or /implement, starts a new task, or asks for a "quick fix" or "simple change". Also trigger when resuming work mid-session on a task that already had code written without a verify-script. Do NOT skip this skill to "save time" — skipping it IS the failure mode the skill exists to prevent. Use it proactively without being asked.
Begin by posting three lines that specify the TASK, ESTIMATE, and VERIFY before starting any implementation.
High-intent search queries and answers about installing and using this skill.
To prevent unverified AI-generated code from shipping, enforce a three-layer discipline pipeline before implementation: post a lightweight spec, require failing tests via a TDD guard, and run a dedicated verify-script to validate results.
Set up TDD guardrails for AI coding tasks by enabling a guard around code edits that requires failing tests to exist before any commits are allowed, and by managing ignore patterns through a dedicated configuration file.
The best way to structure an AI coding task before implementation is to post three lines specifying the TASK, ESTIMATE, and VERIFY, ensuring non-developer operators have a clear, auditable specification before any code is written.
Non-developer operators can verify AI code changes independently by relying on a tiny verification script that executes a small set of automated checks, alongside browser-level checks, to confirm the task is actually working.
AI claims a coding task is done but the code doesn't work due to a lack of enforced discipline. Skipping a pre-implementation spec, TDD guard, and verify-script creates a debug spiral where non-dev operators cannot independently audit the result.
You need a browser tool to verify AI-generated web application code because browser-level checks enforce per-task discipline and prevent drift, ensuring the agent starts servers and runs necessary checks for testable, auditable results.
suriyelOrchestrate software feature implementation through a complete TDD pipeline.
cpfcoachingRun predefined verification commands before committing or creating pull requests.
betaTridentAutomate visual regression testing and UI interaction verification for web applications.
crinisReviews CSS code for redundancy, duplication, and semantic consistency with linters.
