fill-claude-md

Completes and trims CLAUDE.md files by applying content-economics tests to scaffolded stubs.

Updated May 25, 2026
One-click install
npx skills add https://github.com/DATA-AI-XYZ/Tandem --skill fill-claude-md-data-ai-xyz
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: fill-claude-md
Source: https://github.com/DATA-AI-XYZ/Tandem/tree/main/skills/fill-claude-md
Command: npx skills add https://github.com/DATA-AI-XYZ/Tandem --skill fill-claude-md-data-ai-xyz

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? CLAUDE.md files scaffolded by automation contain unverified [auto — verify] markers and <fill in> placeholders, while hand-written files tend to bloat with discoverable or duplicated content that wastes context every session. This Skill applies the judgement layer automation cannot: deciding which non-discoverable, project-specific lines earn a place and removing everything else. ## Core Features & Use Cases - Stub completion: Verifies [auto — verify] command lines against the actual project manifest, confirms or fixes them, and strips the markers. - Content-economics filtering: Runs every candidate line through four tests (broad relevance, non-discoverability, project-specificity, reference-don't-duplicate) plus a deletion gate before it stays. - Trimming mode: Audits existing bloated CLAUDE.md files in reverse, recommending deletions and pointer replacements for duplicated templates or schemas. - Use Case: After running pm:claude-scaffold on a monorepo, use this Skill to confirm the auto-detected test commands, add repo-specific gotchas like "migrations run from repo root, not /services/*", and re-audit until no file reports incomplete. ## Quick Start Fill in the incomplete CLAUDE.md stubs reported by pm:claude-audit and trim any lines that fail the content-economics tests.

Frequently Asked Questions about fill-claude-md

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fill in a CLAUDE.md file after scaffolding?

Run pm:claude-audit with --json to find files marked incomplete, then verify each [auto — verify] command line against the project manifest. Add non-discoverable gotchas and conventions outside the managed markers, strip confirmed markers, and re-audit.

What content belongs in a CLAUDE.md file?

Only lines that pass four tests: broad relevance at that scope, not discoverable by exploring the code, project-specific rather than reusable expertise, and not duplicating content that has a canonical home elsewhere. If deleting a line breaks nothing, delete it.

Should I use @import to share rules across CLAUDE.md files?

No. An @import pulls the target into always-loaded context, re-bloating the root file. Use a plain-text signpost instead, such as "when editing X, read x/CLAUDE.md", so Claude Code loads folder files on demand.

Why does my CLAUDE.md keep getting overwritten by the scaffold?

Content inside PM-KIT:BEGIN/END managed blocks is rewritten by pm:claude-scaffold on each run. Write human prose like gotchas and conventions outside those markers; only the scaffold-owned command lines belong inside.

When should a rule go in a subdirectory CLAUDE.md instead of root?

Place rules by cost-of-a-miss. Root holds rules relevant to almost every task; folder-scoped files hold additive rules true only in that area; thin pointer files reference rules whose home is elsewhere. Subdirectory files must never copy root content.