Generate CLAUDE.md Skill

Produce a project-specific CLAUDE.md through a 4-phase pipeline: SCAN repo facts, DETECT domain enrichment, GENERATE from template, VALIDATE output. The goal is a CLAUDE.md that makes new Claude sessions immediately productive by documenting only verified, project-specific facts.

This skill generates new CLAUDE.md files. It cannot improve an existing one (use claude-md-improver for that), cannot document private dependencies or encrypted configs it cannot read, cannot infer runtime behavior from static files, and cannot replace deep domain expertise -- enrichment patterns are templates, not knowledge.

This skill does not use context: fork because it requires interactive user gates (confirmation when CLAUDE.md already exists, review of generated output), which a forked context would bypass.

Instructions

Execute all phases sequentially. Verify each gate before advancing. Load the template from ${CLAUDE_SKILL_DIR}/references/CLAUDEMD_TEMPLATE.md before Phase 3.

On explicit user request, two optional modes are available:

• Subdirectory CLAUDE.md: Generate per-package CLAUDE.md files for monorepos.
• Minimal Mode ("minimal claude.md"): Only 3 sections -- Overview, Commands, Architecture.

Phase 1: SCAN

Goal: Gather facts about the repository -- language, build system, directory structure, test patterns, config approach.

Step 1: Check for existing CLAUDE.md

ls -la CLAUDE.md .claude/CLAUDE.md 2>/dev/null

If a CLAUDE.md already exists, write output to CLAUDE.md.generated and show a diff, because overwriting a hand-tuned CLAUDE.md destroys work. Inform the user: "CLAUDE.md already exists. Output will be written to CLAUDE.md.generated so you can compare." Continue with all phases -- the generated file is still useful for comparison.

If no CLAUDE.md exists, set output path to CLAUDE.md.

Step 2: Detect language and framework

Check root directory for language indicators:

Read the detected config file to extract: project name, dependencies, language version. Do not assume standard language patterns apply to this project -- read actual source files before writing any section, because conventions vary even within the same language ecosystem.

For Go projects, also check:

# Read go.mod for module path and Go version
head -5 go.mod

For Node.js projects:

# Read package.json for name, scripts, and key dependencies
cat package.json | head -30

Step 3: Parse build system

Parse the Makefile (or equivalent) for actual build targets rather than guessing commands, because the Makefile IS the source of truth for build commands in most repos and may wrap tools with flags, coverage, or race detection that raw invocations would miss.

# Check for Makefile
ls Makefile makefile GNUmakefile 2>/dev/null

# If Makefile exists, extract targets
grep -E '^[a-zA-Z_-]+:' Makefile 2>/dev/null | head -20

Also check for:

• package.json scripts section
• Taskfile.yml
• justfile
• CI config (.github/workflows/, .gitlab-ci.yml)

Record: build command, test command, lint command, "check everything" command. If no build system is found at all, document the gap rather than inventing commands.

Step 4: Map directory structure

# Top-level directories with purpose indicators
ls -d */ 2>/dev/null

For Go projects, also examine:

# Internal packages
ls internal/ 2>/dev/null
ls cmd/ 2>/dev/null
ls pkg/ 2>/dev/null

Categorize directories by role (source, test, config, docs, build, vendor).

Step 5: Find test patterns

# Detect test framework and patterns
# Go
ls *_test.go 2>/dev/null | head -5
# Node.js
ls *.test.ts *.test.js *.spec.ts *.spec.js 2>/dev/null | head -5
# Python
ls test_*.py *_test.py 2>/dev/null | head -5

Read 1-2 representative test files to identify: test framework, assertion library, mocking approach, naming conventions.

Step 6: Detect configuration approach

# Environment-based config
ls .env.example .env.sample 2>/dev/null
# File-based config
ls config.yaml config.json *.toml *.ini 2>/dev/null
# Go flag-based or env-based
grep -r 'os.Getenv\|flag\.\|viper\.\|envconfig' --include='*.go' -l 2>/dev/null | head -5

Step 7: Detect code style tooling

# Linters and formatters
ls .golangci.yml .eslintrc* .prettierrc* .flake8 pyproject.toml .editorconfig 2>/dev/null

If a linter config exists, read it to extract key rules.

Step 8: Check for license headers

# Look for SPDX headers in source files
grep -r 'SPDX-License-Identifier' --include='*.go' --include='*.py' --include='*.ts' -l 2>/dev/null | head -3

If found, note the license type and header convention.

GATE: Language detected. Build targets identified. Directory structure mapped. Test patterns found (or noted as absent). Config approach documented. Proceed ONLY when gate passes.

Phase 2: DETECT

Goal: Identify domain-specific enrichment sources based on repo characteristics. Auto-detect the repo domain and load domain-specific patterns (sapcc Go conventions, OpenStack patterns, etc.) because generic language knowledge is insufficient for project-specific CLAUDE.md generation.

Step 1: Check for sapcc domain (Go repos)

If Go project detected:

# Check go.mod for sapcc imports
grep -i 'sapcc\|sap-' go.mod 2>/dev/null
# Check for sapcc-specific packages
grep -r 'github.com/sapcc' --include='*.go' -l 2>/dev/null | head -5

If sapcc imports found, load enrichment from go-patterns skill patterns:

• Anti-over-engineering principles
• Error wrapping conventions (fmt.Errorf("...: %w", err))
• must.Return scope rules
• Testing patterns (table-driven tests, assertion libraries)
• Makefile management via go-makefile-maker

Step 2: Check for OpenStack/Gophercloud

grep -i 'gophercloud\|openstack' go.mod 2>/dev/null
grep -r 'gophercloud' --include='*.go' -l 2>/dev/null | head -5

If found, note OpenStack API patterns, Keystone auth, and endpoint catalog usage.

Step 3: Detect database drivers

# Go
grep -E 'database/sql|pgx|gorm|sqlx|ent' go.mod 2>/dev/null
# Node.js
grep -E '"pg"|"mysql"|"prisma"|"typeorm"|"knex"|"drizzle"' package.json 2>/dev/null
# Python
grep -E 'sqlalchemy|django|psycopg|asyncpg' pyproject.toml requirements.txt 2>/dev/null

If found, plan to include Database Patterns section.

Step 4: Detect API frameworks

# Go
grep -E 'gorilla/mux|gin-gonic|chi|echo|fiber|go-swagger' go.mod 2>/dev/null
# Node.js
grep -E '"express"|"fastify"|"koa"|"hono"|"next"' package.json 2>/dev/null
# Python
grep -E 'fastapi|flask|django|starlette' pyproject.toml requirements.txt 2>/dev/null

If found, plan to include API Patterns section.

Step 5: Build enrichment plan

Compile a list of which optional CLAUDE.md sections to include and which domain-specific patterns to apply:

Enrichment Plan:
- [ ] sapcc Go conventions (if sapcc imports detected)
- [ ] OpenStack/Gophercloud patterns (if gophercloud detected)
- [ ] Error Handling section (if Go, Rust, or explicit error patterns)
- [ ] Database Patterns section (if DB driver detected)
- [ ] API Patterns section (if API framework detected)
- [ ] Configuration section (if non-trivial config detected)

GATE: Enrichment sources identified. Domain-specific patterns loaded (or explicitly noted as not applicable). Enrichment plan documented. Proceed ONLY when gate passes.

Phase 3: GENERATE

Goal: Load template, fill sections from scan results and enrichment, write CLAUDE.md. Every section must be derived from actual repo analysis (reading files, parsing configs, checking paths) because guessed content wastes the context window and teaches Claude wrong patterns.

Step 1: Load template

Read ${CLAUDE_SKILL_DIR}/references/CLAUDEMD_TEMPLATE.md for the output structure. Follow its structure exactly because consistent structure means Claude sessions can parse CLAUDE.md predictably across projects.

Step 2: Fill required sections

Fill all 6 required sections from Phase 1 scan results:

Section 1 -- Project Overview: Use project name from config file and a description derived from README.md (first paragraph), go.mod module path, or package.json description. List 3-5 key concepts extracted from directory names and core module names. Extract relevant facts from README (project purpose, key concepts) but reframe for Claude's needs -- README is for GitHub visitors (humans browsing the repo), CLAUDE.md is for Claude sessions (AI working in the codebase), so skip installation guides, badges, and user-facing documentation.

Section 2 -- Build and Test Commands: Use ONLY commands found in Makefile, package.json scripts, or equivalent. Format as table. Include "check everything" command prominently. Include single-test and package-test commands. Never write go test ./... without checking the Makefile first because the project's canonical command may include flags, coverage, or race detection.

Section 3 -- Architecture: Map directory structure from Phase 1 Step 4. Identify key components by reading entry points and core modules. Use absolute directory descriptions, not guesses.

Section 4 -- Code Style: Document linter config findings, import ordering (from reading actual source files), naming conventions (from actual code patterns), and tooling that enforces style. Document CLI commands for linting and formatting because those are what Claude actually uses -- do not include IDE/editor setup (VS Code extensions, launch configs) because CLAUDE.md is read by Claude, not by editors.

Section 5 -- Testing Conventions: Document test framework, assertion library, mocking approach, file naming pattern, and integration test requirements from Phase 1 Step 5.

Section 6 -- Common Pitfalls: Derive from actual codebase analysis. Examples of real pitfalls:

• Build tool quirks (e.g., "Makefile is managed by go-makefile-maker -- do not edit directly")
• Dependency gotchas (e.g., "gophercloud v2 migration incomplete -- some packages still use v1")
• Test requirements (e.g., "integration tests require PostgreSQL running locally")
• Config requirements (e.g., "OS_AUTH_URL must be set for any OpenStack operation")

Do NOT invent pitfalls because fabricated warnings erode trust. If nothing notable was found, include 1-2 based on the build system (e.g., "run make check before committing").

Ban these generic phrases -- they waste tokens and provide zero project-specific signal:

• "use meaningful variable names"
• "write clean code"
• "follow best practices"
• "ensure code quality"
• "maintain consistency"
• "keep it simple"
• "write tests"
• "handle errors properly"

If you find yourself writing generic advice, remove it entirely because leaving a section out is better than filling it with filler.

Step 3: Fill optional sections

Based on the Phase 2 enrichment plan, fill applicable optional sections. If an optional section is included, it must be backed by evidence from repo analysis because optional sections without evidence are worse than omitted sections:

• Error Handling: For Go repos, document wrapping conventions found in source. For sapcc repos, include fmt.Errorf("...: %w", err) pattern and note error checking tools from linter config.
• Database Patterns: Document the driver/ORM, migration tool, and key query patterns found in source.
• API Patterns: Document the framework, auth mechanism, and response format found in source.
• Configuration: Document config source (env vars, files, flags), key variables from .env.example, and override precedence.

Step 4: Apply domain enrichment

For sapcc Go repos (detected in Phase 2 Step 1), integrate these patterns into the relevant sections:

In Code Style, add:

• Anti-over-engineering: prefer simple, readable solutions over clever abstractions
• Scope must.Return to init functions and test helpers only
• Error wrapping: always add context with fmt.Errorf("during X: %w", err)

In Testing Conventions, add:

• Table-driven tests as the default pattern
• Relevant assertion libraries detected in go.mod

In Common Pitfalls, add:

• go-makefile-maker manages the Makefile (if detected)
• Any sapcc-specific patterns found in the codebase

Step 5: Write output

Write the completed CLAUDE.md (or CLAUDE.md.generated) to the output path determined in Phase 1 Step 1. Verify every path mentioned in the output exists and every command is runnable before writing, because a CLAUDE.md with broken paths is worse than no CLAUDE.md -- it teaches Claude to trust wrong information.

If writing to CLAUDE.md.generated, also show the user a summary diff:

diff CLAUDE.md CLAUDE.md.generated 2>/dev/null || echo "New file created"

GATE: CLAUDE.md written. All required sections populated with project-specific content (no placeholders). Optional sections populated based on enrichment plan. Output path is correct. Proceed ONLY when gate passes.

Phase 4: VALIDATE

Goal: Verify the generated CLAUDE.md is accurate, complete, and free of generic filler.

Step 1: Verify all paths exist

Extract every file path and directory path mentioned in the generated CLAUDE.md. Check each one with test -e because "probably exists" is not verified -- one broken path undermines the entire document:

# For each path mentioned in the output
test -e "<path>" && echo "OK: <path>" || echo "MISSING: <path>"

If any path is missing, fix or remove the reference.

Step 2: Verify all commands parse

Extract every command mentioned in the generated CLAUDE.md. Check each one exists:

# For each command mentioned (e.g., "make lint")
# Verify the tool exists
which <tool> 2>/dev/null || echo "MISSING: <tool>"
# For Makefile targets, verify the target exists
grep -q '^<target>:' Makefile 2>/dev/null || echo "MISSING TARGET: <target>"

If a command references a missing tool or target, fix or remove it.

Step 3: Check for remaining placeholders

Search the output for placeholder patterns:

grep -E '\{[^}]+\}|TODO|FIXME|TBD|PLACEHOLDER' <output_file>

If any placeholders remain, fill them from repo analysis or remove the containing section.

Step 4: Check for generic filler

Search the output for banned generic phrases:

- "use meaningful variable names"
- "write clean code"
- "follow best practices"
- "ensure code quality"
- "maintain consistency"
- "keep it simple"
- "write tests"
- "handle errors properly"

If any generic filler is found, replace with project-specific content or remove.

Step 5: Report summary

Display a validation report:

CLAUDE.md Generation Complete
==============================
Output: <path>
Sections: <count> required + <count> optional
Paths verified: <count> OK, <count> fixed
Commands verified: <count> OK, <count> fixed
Placeholders: <count> (should be 0)
Generic filler: <count> (should be 0)

Domain enrichment applied:
- <enrichment 1>
- <enrichment 2>

Next steps:
- Review the generated file
- If CLAUDE.md.generated: compare with existing CLAUDE.md and merge manually
- Use /claude-md-improver to refine further

GATE: All paths resolve. All commands verified. No placeholders remain. No generic filler detected. Validation report displayed.

Examples

Example 1: Go sapcc Repository

User says: "generate claude.md" Actions:

1. SCAN: Detect go.mod, parse Makefile targets (make build, make check, make lint), map cmd/, internal/, pkg/ directories, find _test.go files with testify assertions
2. DETECT: Find github.com/sapcc imports in go.mod, load sapcc conventions (anti-over-engineering, error wrapping, go-makefile-maker)
3. GENERATE: Fill template with Go-specific content, add sapcc enrichment to Code Style and Testing sections, include error handling section
4. VALIDATE: Verify all internal/ paths exist, confirm make check target exists in Makefile, no placeholders Result: CLAUDE.md with sapcc-aware conventions, real Makefile commands, verified paths

Example 2: Node.js/TypeScript Project

User says: "create a claude.md for this repo" Actions:

1. SCAN: Detect package.json, extract npm scripts (npm test, npm run build, npm run lint), map src/, tests/ directories, find .test.ts files with vitest
2. DETECT: Find express in dependencies, plan API Patterns section, no domain enrichment
3. GENERATE: Fill template with TypeScript content, include API patterns (Express routes, middleware), testing conventions (vitest, co-located tests)
4. VALIDATE: Verify all paths, confirm npm scripts exist, no generic filler Result: CLAUDE.md with actual npm commands, Express API patterns, vitest testing conventions

Example 3: Existing CLAUDE.md

User says: "generate claude.md" Actions:

1. SCAN: Find existing CLAUDE.md, set output to CLAUDE.md.generated, continue analysis
2. DETECT: Standard detection, no special domain
3. GENERATE: Write to CLAUDE.md.generated
4. VALIDATE: Show diff between existing and generated, suggest using claude-md-improver to merge Result: CLAUDE.md.generated alongside existing file, with diff for comparison

Error Handling

Error: No Build System Detected

Cause: No Makefile, package.json scripts, Taskfile, or other build configuration found. Solution: Generate a minimal CLAUDE.md documenting only what can be verified (directory structure, language, test patterns). Note prominently in the Build and Test Commands section: "No build system detected -- add build commands manually." Continue with all other phases.

Error: CLAUDE.md Already Exists

Cause: Repository already has a CLAUDE.md (root or .claude/ directory). Solution: Write output to CLAUDE.md.generated. Show diff between existing and generated files. Suggest using claude-md-improver to merge improvements. Never overwrite without explicit user confirmation.

Error: Unknown Language

Cause: No recognized language indicator files in the repository root. Solution: Produce a language-agnostic CLAUDE.md focusing on directory structure, Makefile targets (if present), and any README content. Note the gap: "Language could not be auto-detected -- add language-specific sections manually."

References

Reference Files

• ${CLAUDE_SKILL_DIR}/references/CLAUDEMD_TEMPLATE.md: Template structure for generated CLAUDE.md files with required and optional sections
• Official Anthropic claude-md-management:claude-md-improver: Companion skill for improving existing CLAUDE.md files (use after generation for refinement)

Companion Skills

• go-patterns: Domain-specific patterns for sapcc Go repositories (loaded during Phase 2 enrichment)
• codebase-overview: Deeper codebase exploration when CLAUDE.md generation needs more architectural context