claude-code.md

Installing Results-Driven (以终为始) for Claude Code

Recommended: Run ./install.sh from the repo root for automated installation with dual-mode support (@ reference + inline fallback). The manual steps below are for reference or troubleshooting.

Quick install

# Clone the repo
git clone https://github.com/d-wwei/results-driven.git
cd results-driven

# Copy skill files
mkdir -p ~/.claude/skills/results-driven
cp SKILL.md anti-patterns.md examples.md ~/.claude/skills/results-driven/

# Inject core rules into CLAUDE.md (direct content injection — works on all versions)
cat cognitive-protocol.md >> ~/.claude/CLAUDE.md

How it works in Claude Code

Claude Code reads ~/.claude/CLAUDE.md at the start of every conversation. The core cognitive rules are directly injected into CLAUDE.md, so they load automatically — no manual trigger needed, no version-specific features required.

The full SKILL.md in ~/.claude/skills/results-driven/ serves as a detailed reference that Claude Code can access when deeper guidance is needed.

Two layers

Layer	File	Loaded when	Purpose
Core rules	~/.claude/CLAUDE.md (appended)	Every conversation (part of CLAUDE.md)	Always-on completion integrity
Full reference	~/.claude/skills/results-driven/SKILL.md	When agent detects relevance	Detailed anti-patterns, examples, composition guide

Verifying installation

Start a new Claude Code conversation and give a multi-part task (e.g., "Add input validation, error messages, and unit tests to this form"). The response should:

• Address all three parts, not just the easiest ones
• Show evidence that changes work (test output, verification)
• Not claim "Done!" until all parts are delivered

Uninstalling

# Remove the Results-Driven section from ~/.claude/CLAUDE.md
# (search for the "# Results-Driven (以终为始) — Cognitive Protocol" header and delete that section)

# Remove skill files
rm -rf ~/.claude/skills/results-driven