tierward

import fs from 'fs-extra'; import path from 'path';

const tierDescriptions = { s: 'Fast Lane (Tier S) - minimal pipeline for bugfixes and small changes', m: 'Standard (Tier M) - 5-phase pipeline for feature blocks', l: 'Full (Tier L) - complete governance pipeline for complex, long-running projects', };

export async function generateReadme(config, targetDir) { const tier = (config.tier || 's').toLowerCase(); const tierDesc = tierDescriptions[tier] || '';

const content = `# ${config.projectName}

${config.description}

Development

```bash ${config.installCommand || 'npm install'} ${config.devCommand || 'npm run dev'} ```

Testing

```bash ${config.testCommand || 'npm test'} ${config.typeCheckCommand ? config.typeCheckCommand + ' # type check' : ''} ```

AI-Assisted Development

This project uses Claude Code with a governance scaffold provided by tierward.

Pipeline: ${tierDesc}

Key files:

• `CLAUDE.md` - project context for Claude (tech stack, conventions, known patterns)
• `.claude/rules/pipeline.md` - mandatory development workflow
• `.claude/settings.json` - tool permissions and governance hooks
• `docs/adr/` - architecture decision records

Every AI-assisted commit is tagged via `attribution.commit` in `.claude/settings.json` and should be reviewed by a human before merging to shared branches.

Getting started

• Install Claude Code CLI
• Open the project in your terminal and run `claude`
• Claude will load `CLAUDE.md` and `.claude/rules/` automatically

Governance

Control	How it works
Test gate	Claude cannot complete a task until `${config.testCommand
Audit log	Every tool use is logged to `~/.claude/audit/`
Secret scanning	Pre-commit hook blocks accidental credential commits
CODEOWNERS	`.claude/` directory requires tech lead review on all PRs

Contributing

See `.github/PULL_REQUEST_TEMPLATE.md` for the PR checklist. All PRs that include AI-generated code must complete the AI review checklist. `;

await fs.writeFile(path.join(targetDir, 'README.md'), content); }