config-audit/commands/discover.md

name	description	argument-hint	allowed-tools	model
config-audit:discover	Phase 1 - Initialize session, auto-detect scope, and discover config files	[current|repo|home|full] [--delta]	Read, Write, Edit, Glob, Grep, Agent, AskUserQuestion, Bash	opus

Config-Audit: Discover (Phase 1)

Initialize a new audit session and discover all Claude Code configuration files.

Usage

/config-audit discover              # Auto-detect scope
/config-audit discover current      # Force current directory scope
/config-audit discover repo         # Force git repository scope
/config-audit discover home         # Force home/global scope
/config-audit discover full         # Force full machine scope
/config-audit discover --delta      # Incremental re-scan (changed files only)

Implementation

Step 1: Initialize session and greet

Generate session ID (YYYYMMDD_HHmmss), create directories:

mkdir -p ~/.claude/config-audit/sessions/{session-id}/findings 2>/dev/null

Step 2: Determine scope

If the user provided a scope argument, use it. Otherwise, auto-detect:

1. Run git rev-parse --show-toplevel 2>/dev/null
2. If inside a git repo → repo scope
3. If pwd is $HOME → home scope
4. Otherwise → current directory scope

Tell the user:

## Configuration Discovery

**Scope:** {Repository|Home|Current directory|Full machine} — `{path}`
Finding all Claude Code configuration files (CLAUDE.md, settings, hooks, rules, MCP servers)...

Step 3: Resolve paths

Scope	What gets scanned
current	Current directory + parent CLAUDE.md files up to root + ~/.claude/
repo	Git repo root + ~/.claude/
home	~/.claude/ only
full	~/.claude/ (depth 10), managed paths, all dev dirs under $HOME

Step 4: Delta mode (if --delta)

If --delta flag:

1. Find previous baseline from ~/.claude/config-audit/sessions/*/discovery.json
2. If no previous: "No previous scan found. Running full discovery instead."
3. Compare file mtimes/sizes to classify as changed/new/deleted/unchanged
4. Only scan changed + new files

Step 5: Run discovery

Run the scan orchestrator silently to discover and scan files. Default mode emits humanized JSON — each finding in scan-results.json carries userImpactCategory, userActionLanguage, and relevanceContext alongside the v5.0.0 fields. Pass --raw through if the user requested it (produces v5.0.0 verbatim envelope; humanizer fields absent).

RAW_FLAG=""
if echo "$ARGUMENTS" | grep -q -- "--raw"; then RAW_FLAG="--raw"; fi
# Set to the flag itself for full/home scope, otherwise leave empty. Never pass
# a placeholder wrapped in square brackets: it does not start with a dash, so
# the orchestrator's arg loop takes it as the SCAN TARGET and silently scans a
# path that does not exist.
SCOPE_FLAGS=""   # e.g. SCOPE_FLAGS="--full-machine" or SCOPE_FLAGS="--global"
node ${CLAUDE_PLUGIN_ROOT}/scanners/scan-orchestrator.mjs <target-path> --output-file ~/.claude/config-audit/sessions/{session-id}/findings/scan-results.json $SCOPE_FLAGS $RAW_FLAG 2>/dev/null; echo $?

Check exit code: 0/1/2 → normal. 3 → "Discovery encountered an error. Try a narrower scope."

Step 6: Save scope and state

Write scope.yaml and state.yaml to session directory. Update state with all four fields .claude/rules/state-management.md requires: current_phase: "discover", completed_phases: [discover], next_phase: "analyze", and updated_at. The last two are what make an interrupted run resumable.

Step 7: Present summary

Read the scan results file using the Read tool. When you surface initial findings, group them by userImpactCategory and lead each line with userActionLanguage rather than raw severity prefiks — the humanizer already mapped severity to plain-language phrasing ("Fix this now", "Fix soon", "Fix when convenient", "Optional cleanup", "FYI") so the rest of the toolchain sees consistent wording.

Full scan:

### Discovery Complete

**{scope_type}** scope — found {total_files} configuration files:

| Type | Count |
|------|-------|
| CLAUDE.md | {n} |
| Settings | {n} |
| MCP configs | {n} |
| Rules | {n} |
| Hooks | {n} |
| Other | {n} |

Initial scan found {finding_count} items to review (grouped by impact: {comma-separated counts per userImpactCategory}).

**Next:** Run `/config-audit analyze` to generate your analysis report.

Delta scan:

### Delta Discovery Complete

Compared against baseline from {previous-session-id}:

| Status | Files |
|--------|-------|
| Changed | {n} |
| New | {n} |
| Deleted | {n} |
| Unchanged | {n} |

Only {changed+new} file(s) scanned (vs {total} full scan).

**Next:** Run `/config-audit analyze` to generate your analysis report.

Config File Patterns

Pattern	Description
**/CLAUDE.md	Project instructions
**/CLAUDE.local.md	Local overrides
**/.claude/settings.json	Project settings
**/.mcp.json	MCP servers
**/.claude/rules/*.md	Modular rules

For global: ~/.claude/CLAUDE.md, ~/.claude/settings.json, ~/.claude.json, ~/.claude/agents/*.md

Error Handling

• If scanner fails, report to user in plain language and suggest narrower scope
• If path doesn't exist, tell user and suggest alternatives
• If git command fails for repo scope, silently fall back to current
• If no config files found, explain: "No Claude Code configuration files found. Start with /config-audit feature-gap to see what's recommended."