Your CLAUDE.md is on every prompt. Every line is a tax.
Paste your CLAUDE.md, AGENTS.md, .cursorrules, or .grokrules and see which lines are silently bloating your context. Token math, Karpathy-12 scorecard, skill recommendations, zero signup. Drift monitoring on the paid tier.
.cursorrules
- L5aspirational
Absolute "always" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L7vague
"good" has no testable meaning; the agent cannot tell when it succeeded.
→ replace with a concrete rule (a number, a named tool, a forbidden pattern).
- L7aspirational
Absolute "always" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L8vague
"appropriately" has no testable meaning; the agent cannot tell when it succeeded.
→ replace with a concrete rule (a number, a named tool, a forbidden pattern).
- L8aspirational
Absolute "always" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L9aspirational
Absolute "always" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L10vague
"carefully" has no testable meaning; the agent cannot tell when it succeeded.
→ replace with a concrete rule (a number, a named tool, a forbidden pattern).
- L10aspirational
Absolute "always" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L11vague
"best" has no testable meaning; the agent cannot tell when it succeeded.
→ replace with a concrete rule (a number, a named tool, a forbidden pattern).
- L11aspirational
Absolute "always" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L13aspirational
Absolute "never" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L15aspirational
Absolute "never" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L15missing_why
Prohibition with no "why". The agent will follow until edge cases, then guess.
→ add a one-line "Why:" with the constraint or past incident.
- L16aspirational
Absolute "never" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
- L16missing_why
Prohibition with no "why". The agent will follow until edge cases, then guess.
→ add a one-line "Why:" with the constraint or past incident.
- L17missing_why
Prohibition with no "why". The agent will follow until edge cases, then guess.
→ add a one-line "Why:" with the constraint or past incident.
- L29vague
"good" has no testable meaning; the agent cannot tell when it succeeded.
→ replace with a concrete rule (a number, a named tool, a forbidden pattern).
- L29aspirational
Absolute "always" with no escape clause. Real codebases have exceptions.
→ add "unless X" or remove the absolute.
Static analysis only. The free tier runs every check in your browser and never sends the file anywhere. Paid tier adds session-trace analysis (which rules actually fire), continuous drift monitoring on a GitHub webhook, and a per-engineer cost rollup for teams.
Claude Code shipped a memory primitive (CLAUDE.md) and gave it almost no guardrails. Power users figured out the hard way that the file is sent on every turn, that it competes for the same context budget that drives output quality, and that a sloppy 8k-token file silently taxes every prompt for the rest of the session.
every single API call to Claude sends the whole context, including prompts, meaning that all this extra text in CLAUDE.md is sent over and over.— caymanjim, HN 47581701
Then Anthropic split programmatic Claude Code billing into a separate monthly Agent SDK credit pool starting 2026-06-15. A chatty config file is now a literal line item on the credit balance.
Existing tools split into generators (paste a stack, get a starter) and cost meters (tell you what you spent). Neither audits the file you already wrote. ccmd does.
Same rubric across four agent ecosystems. Polyglot from day one so you can audit every config in your repo without learning four tools.
- claude.mdAnthropic Claude Code memory file.
- agents.mdOpenAI Codex / GPT-5.5 agent contract.
- .cursorrulesCursor IDE rules; .mdc multi-file variant supported.
- .grokrulesxAI Grok Build CLI (May 2026). Native parsing.
Andrej posted twelve rules in January. The community measured a 41% → 11% mistake-rate drop.
ccmd's default rubric scores your file against all twelve: Think-Before-Coding, Simplicity-First, Surgical-Changes, Goal-Driven-Execution, Avoid-Silent-Assumptions, No-Orthogonal-Damage, Tests-as-Truth, Concise-Output, Stack-Awareness, Tool-Preference, Failure-Mode-Coverage, Self-Improvement-Loop. The hero panel above already tells you your score; read the long version.
- · Paste-and-score, no signup
- · Karpathy-12 rubric
- · Token math, dollar estimate
- · Copy optimized version
- · 100% in-browser
- · Everything in free
- · GitHub webhook on push
- · Weekly drift email
- · Session-trace cost analysis
- · Skill recommendations from skillhu.bz
- · Everything in solo
- · Shared CLAUDE.md dashboard
- · PR diff comments on config
- · Per-engineer cost attribution
- · Quarterly polyglot rollup
Why not just ask Claude to optimize my CLAUDE.md?
Claude pasting its own file back at itself is a 200-token tip. ccmd measures: which lines fire, which never do, which contradict the system prompt, and what every line costs at Opus 4.7 rates. A model cannot count its own tokens reliably and cannot see your billing.
Does it work with Codex, Cursor, Grok Build?
Yes. Polyglot from day one. AGENTS.md (Codex), .cursorrules (Cursor IDE), and .grokrules (xAI Grok Build, shipped May 2026) all use the same scoring rubric. One tool for every config file in your repo.
How is this different from Anthropic's claude-code-setup plugin?
claude-code-setup recommends structure for a brand-new project. ccmd grades the file you already have, against a rubric tied to specific past failure modes, with a token-cost number on every finding.
Will the June 15 Agent SDK billing change make agents more expensive?
Yes for programmatic use. Pro $20 / Max 5x $100 / Max 20x $200 separate monthly credit pool at full API rates, no rollover. ccmd flags the chatty patterns in your CLAUDE.md that disproportionately burn programmatic credits (Read-heavy loops, large file reads in tight cycles, redundant Bash steps).