New read-only command that shows everything Claude Code actually loads for a given repo — plugins, skills, MCP servers, hooks, CLAUDE.md cascade — with source attribution (user/project/plugin) and rough token estimates. Helps identify candidates for disabling without guessing. Added: - scanners/lib/active-config-reader.mjs — pure async helper: readActiveConfig, detectGitRoot, walkClaudeMdCascade, readClaudeJsonProjectSlice (longest-prefix matching for .claude.json projects), enumeratePlugins, enumerateSkills, readActiveHooks, readActiveMcpServers, estimateTokens (markdown 4 c/tok, json 3.5 c/tok, frontmatter cap 150 tokens, item flat 15) - scanners/whats-active.mjs — thin CLI shim: --json, --output-file, --verbose, --suggest-disables - commands/whats-active.md — renders tables via Read tool; honors UX rules - tests/lib/active-config-reader.test.mjs — 36 tests, all green (integration fixture built in tmpdir with fake HOME, .claude.json prefix matching, plugin discovery, hook/MCP merge from all scopes) Verified: - Performance budget: <2s wall-clock (smoke test: 102ms on real repo) - Token estimates within ±20% of hand-computed values - Read-only: no writeFile/mkdir/unlink in production code - Self-audit: Plugin Health scanner reports 0 findings (Grade A) - Full test suite: 522 tests, 512 pass (10 pre-existing conflict-detector failures on main — unrelated to this change, reproducible on clean HEAD) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
6.4 KiB
| name | description | argument-hint | allowed-tools | model |
|---|---|---|---|---|
| config-audit:whats-active | Show which plugins, skills, MCP servers, hooks, and CLAUDE.md files are active for a repo — with token estimates | [path] [--json] [--verbose] [--suggest-disables] | Read, Glob, Bash | sonnet |
Config-Audit: What's Active
Show a complete, read-only inventory of everything Claude Code loads for a given repo — plugins, skills, MCP servers, hooks, CLAUDE.md cascade — with source attribution and rough token estimates. Helps identify candidates for disabling without guessing.
UX Rules (MANDATORY — from .claude/rules/ux-rules.md)
- Never show raw JSON or stderr output. Always use
--output-file+2>/dev/null. - Narrate before acting. Tell the user what you're about to do.
- Read, don't dump. Read the JSON file and render formatted tables.
- End with context-sensitive next steps.
Implementation
Step 1: Parse $ARGUMENTS
Split $ARGUMENTS into a path and flags. Path is the first non-flag argument. Default to . (current working directory). Recognized flags:
--json— emit raw JSON instead of rendered tables (power-user mode)--verbose— include per-file byte/line detail--suggest-disables— append deterministic disable-candidates + LLM-judgment pass
Step 2: Run the CLI silently
Tell the user: "Reading active configuration for <path>..."
TMPFILE="/tmp/ca-whats-active-$$.json"
node ${CLAUDE_PLUGIN_ROOT}/scanners/whats-active.mjs <path> --output-file "$TMPFILE" [--verbose] [--suggest-disables] 2>/dev/null; echo $?
Exit code handling:
0→ continue3→ tell user: "Couldn't read configuration. Check that the path exists and is a directory." Stop.
Step 3: If --json was requested, cat the file and stop
cat "$TMPFILE"
Do NOT render tables in JSON mode.
Step 4: Read JSON and render
Use the Read tool on $TMPFILE. Extract:
meta.repoPath,meta.durationMs,meta.gitRoot,meta.projectKeytotals.estimatedTokens.grandTotal(and subtotals)claudeMd.files[]— render cascade tableplugins[]— render plugin tableskills[]— render skills tablemcpServers[]— render MCP table (disabled shown italic)hooks[]— render hooks table
Render as markdown:
**Active configuration for `<repoPath>`** — ~{grandTotal} tokens loaded at startup
{if gitRoot != repoPath: "Git root: `<gitRoot>`"}
{if projectKey: "`.claude.json` project slice: `<projectKey>`"}
### CLAUDE.md cascade ({claudeMd.files.length} files, ~{claudeMd.estimatedTokens} tokens)
| Scope | Path | Bytes | Lines |
|-------|------|-------|-------|
| {scope} | `<path>` | {bytes} | {lines} |
| ... | ... | ... | ... |
### Plugins ({plugins.length}, ~{plugins subtotal} tokens)
| Plugin | Version | Commands | Agents | Skills | Hooks | Rules | Tokens |
|--------|---------|----------|--------|--------|-------|-------|--------|
| {name} | {version} | {commands} | {agents} | {skills} | {hooks} | {rules} | ~{estimatedTokens} |
### Skills ({skills.length}, ~{skills subtotal} tokens)
| Skill | Source | Tokens |
|-------|--------|--------|
| {name} | {source}{if pluginName: ` (${pluginName})`} | ~{estimatedTokens} |
### MCP Servers ({mcpServers.length}, ~{mcpServers subtotal} tokens)
| Server | Source | Status | Command |
|--------|--------|--------|---------|
| {name} | {source} | {enabled ? "enabled" : "*disabled*"} | `{command}` |
### Hooks ({hooks.length}, ~{hooks subtotal} tokens)
| Event | Matcher | Source |
|-------|---------|--------|
| {event} | {matcher or "-"} | {source} |
### Settings cascade
| Scope | Path | Keys |
|-------|------|------|
| user | `<path>` | {keyCount} |
| project | `<path>` | {keyCount} |
| local | `<path>` | {keyCount or "(missing)"} |
### Totals
| Category | Items | Estimated tokens |
|----------|-------|------------------|
| CLAUDE.md | {claudeMdFiles} | ~{claudeMd} |
| Plugins | {plugins} | ~{plugins} |
| Skills | {skills} | ~{skills} |
| MCP servers | {mcpServers} | ~{mcpServers} |
| Hooks | {hooks} | ~{hooks} |
| **Grand total** | — | **~{grandTotal}** |
_Estimates assume ~4 chars/token (Claude ballpark). Real token count varies ±15%._
Step 5: If --verbose, add per-file detail
For each CLAUDE.md file, skill, and plugin, include a nested "Details" list with bytes, lines, and full path.
Step 6: If --suggest-disables, show candidates
First show deterministic signals from suggestDisables.candidates[]:
### Disable candidates (deterministic)
| Kind | Name | Reason | Confidence |
|------|------|--------|------------|
| {kind} | {name} | {reason} | {confidence} |
Then run LLM judgment — check git log --oneline -20 and project manifests (package.json/Cargo.toml/etc.) to propose up to 3 additional candidates. For each candidate, you MUST:
- Name the specific redundancy
- Name the signal the user should check to confirm
Do NOT suggest items you can't name concrete redundancy for. If you can't find 3 strong candidates, return fewer or zero.
Step 7: Cleanup and next steps
rm -f "$TMPFILE"
### What's next
- **`/config-audit posture`** — check configuration health (A-F grades per area)
- **`/config-audit feature-gap`** — context-aware recommendations for features you aren't using
- **Disable a plugin:** edit `~/.claude/settings.json` → `enabledPlugins` (remove the entry)
- **Disable an MCP server:** edit `~/.claude.json` → `projects.<path>.disabledMcpjsonServers`
- **Re-run with flags:** `/config-audit whats-active --verbose` (details) or `--suggest-disables` (pruning help)
Scope and limits
- Read-only. This command never writes to configuration files — no mkdir, no edits, no deletes.
- Single repo. Scans one repo path per invocation. Cross-repo rollups are out of scope.
- Ballpark token counts. Estimates are deterministic but not calibrated against Claude's tokenizer. Use them to compare categories, not to predict exact billing.
- No runtime queries. We inspect config files only — we do not connect to MCP servers or invoke hooks.
Error handling
| Condition | Action |
|---|---|
| Exit code 3 | Tell user path is invalid, suggest checking path exists |
| JSON parse fails (shouldn't happen — CLI writes valid JSON) | Tell user to re-run, mention this as a bug to report |
| No plugins, no CLAUDE.md, no hooks found | Still render with zeroes; suggest /config-audit feature-gap for setup help |