xray

name: xray description: X-ray deep scan — inspect Claude Code auto-memory, installed skills, or session history. Use when user says "xray", "x-ray", "memory", "scan memory", "what do you remember", "show memories", "memory stats", "forget", "list skills", "installed skills", "session history", or wants to inspect what the AI remembers across sessions. Do NOT trigger for Oracle vault/ψ (use /trace) or session handoffs (use /inbox). argument-hint: "[memory|skills|sessions]"

/xray - X-Ray Deep Scan

Inspect and manage Claude Code auto-memory, installed skills, and session history.

Step 0 — Anchor (date-stamp + root)

date "+🕐 %H:%M %Z (%A %d %B %Y)"

# Find oracle root — git toplevel that has CLAUDE.md + ψ/
ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null)
if [ -n "$ORACLE_ROOT" ] && [ -f "$ORACLE_ROOT/CLAUDE.md" ] && { [ -d "$ORACLE_ROOT/ψ" ] || [ -L "$ORACLE_ROOT/ψ" ]; }; then
  PSI="$ORACLE_ROOT/ψ"
elif [ -f "$(pwd)/CLAUDE.md" ] && { [ -d "$(pwd)/ψ" ] || [ -L "$(pwd)/ψ" ]; }; then
  ORACLE_ROOT="$(pwd)"
  PSI="$ORACLE_ROOT/ψ"
else
  echo "⚠️ Not in oracle repo (no CLAUDE.md + ψ/ at git root). Reading from pwd."
  ORACLE_ROOT="$(pwd)"
  PSI="$ORACLE_ROOT/ψ"
fi

Subcommands

Target: memory (default)

Memory Location

ENCODED=$(pwd | sed 's|^/|-|; s|[/.]|-|g')
MEMORY_DIR="$HOME/.claude/projects/${ENCODED}/memory"

Usage

/xray                     # Default: scan memory — list all memories with types
/xray memory              # Same as above
/xray memory scan         # Same as above
/xray memory read <name>  # Read a specific memory file
/xray memory stats        # Show counts by type, total size, age
/xray memory types        # Group memories by type
/xray memory clean        # Find stale/outdated memories
/xray memory forget <name># Archive a memory (after confirmation)
/xray skills              # List installed Oracle skills
/xray sessions            # Show session history

Mode 1: Scan (default)

List all memory files with type, age, and description.

MEMORY_DIR="$HOME/.claude/projects/$(pwd | sed 's|^/|-|; s|[/.]|-|g')/memory"

Read each .md file (except MEMORY.md), extract frontmatter:

For each file, display:

🧠 Memory Scanner — [project name]

  Type        Name                              Age     Description
  ─────────── ───────────────────────────────── ─────── ────────────────────────
  feedback    never_push_main_skills            2d      Always use branch + PR
  feedback    talk_to_not_oracle_thread         9d      Use /talk-to for threads
  reference   maw_cli                           9d      maw hey, peek, ls commands
  reference   pulse_cli                         9d      board, timeline, scan
  user        work_patterns                     9d      maker's schedule, peak hours

  Total: 14 memories (5 feedback, 3 reference, 1 user, 0 project)
  Index: MEMORY.md (60 lines)

How to parse

For each .md file in MEMORY_DIR (excluding MEMORY.md):

1. Read frontmatter between --- markers
2. Extract name, description, type fields
3. Get file age from modification time: stat -c %Y or date -r
4. Format age as: Xd (days), Xh (hours), Xm (minutes)

Sort by type, then by age (newest first).

Mode 2: Read

/xray memory read <name>

Find file matching <name> (partial match OK):

ls "$MEMORY_DIR"/*<name>*.md | head -1

Display full content with frontmatter highlighted.

Mode 3: Stats

/xray memory stats

🧠 Memory Stats — [project name]

  Memories:    14 files
  Index:       MEMORY.md (60 lines)
  Total size:  12.4 KB
  Oldest:      9 days (reference_maw_cli.md)
  Newest:      2 days (feedback_never_push_main.md)

  By type:
    feedback   ████████████  5  (36%)
    reference  ████████      3  (21%)
    user       ████          1  (7%)
    project    ░░░░          0  (0%)

  Top keywords: oracle, push, maw, agent, fleet

How to calculate

• Count files by type (from frontmatter)
• Sum file sizes
• Extract keywords: split all description fields into words, count frequency, show top 5

Mode 4: Types

/xray memory types

Group and display memories by type:

🧠 Memory Types

  📝 feedback (5)
    - never_push_main_skills — Always use branch + PR
    - talk_to_not_oracle_thread — Use /talk-to for threads
    - use_workon_skill — Use skill flow, don't edit directly
    - worktree_naming — worktree name = repo name
    - never_maw_done_without_approval — Ask before cleanup

  📚 reference (3)
    - maw_cli — maw hey, peek, ls commands
    - pulse_cli — board, timeline, scan
    - office_hash_routes — all views, deploy flow

  👤 user (1)
    - work_patterns — maker's schedule, peak hours

  📋 project (0)
    (none)

Mode 5: Clean

/xray memory clean

Find potentially stale memories:

1. Read each memory file
2. If type: reference and mentions file paths → check if files still exist
3. If type: project → check if still relevant (age > 30 days = likely stale)
4. Show candidates:

🧹 Memory Clean — candidates

  ⚠️  reference_office_hash_routes.md (9d)
      Mentions file paths — verify still current

  ✓  All other memories look current

  Run '/xray memory forget <name>' to archive a specific memory.
  Nothing is deleted — archived for history.

NEVER auto-delete. Only suggest. User decides. Stale memories are archived, not deleted.

Mode 6: Forget (Archive)

/xray memory forget <name>

"Nothing is Deleted" — contradicted or stale memories are archived, not destroyed.

1. Find the file
2. Show its content
3. Ask confirmation: "Archive this memory? (yes/no)"
4. If yes:
   • Create archive dir: mkdir -p <MEMORY_DIR>/archive/
   • Move the file: mv <MEMORY_DIR>/<file>.md <MEMORY_DIR>/archive/<file>.md
   • Remove its entry from MEMORY.md
   • Confirm: "Archived: → archive/.md"
5. If no: "Kept: "

IMPORTANT: Never rm a memory file. Always mv to archive/.

Cross-Project View

If user says /xray memory scan --all or /xray memory stats --all:

ls -d "$HOME/.claude/projects"/*/memory/ 2>/dev/null

Show memory counts per project:

🧠 All Projects

  Project                          Memories  Size
  ──────────────────────────────── ──────── ──────
  neo-oracle                       14        12.4KB
  arra-oracle-skills-cli                 8         6.2KB
  maw-js                            5         3.1KB
  office-8bit                       3         2.0KB

  Total: 30 memories across 4 projects

Target: skills

/xray skills

List all installed Oracle skills with their profiles and status.

arra-oracle-skills list -g

Display:

🔬 Installed Skills

  Skill                Profile     Status
  ──────────────────── ─────────── ──────
  forward              seed        ✓ installed
  retrospective        seed        ✓ installed
  rrr                  seed        ✓ installed
  awaken               standard    ✓ installed
  trace                standard    ✓ installed
  learn                standard    ✓ installed

  Total: N skills installed
  Profile: [seed|standard|full]

How to gather

1. Run arra-oracle-skills list -g to get installed skills
2. Cross-reference with profiles from profiles.ts
3. Show which profile each skill belongs to

Target: sessions

/xray sessions

Show recent Claude Code session history and retrospectives.

1. Look for retrospectives in $PSI/memory/retrospectives/
2. Look for session logs in $PSI/memory/logs/
3. Display recent sessions with dates and summaries:

📜 Session History

  Date         Duration  Summary
  ──────────── ──────── ────────────────────────────────
  2026-03-25   45min     Built xray skill, updated profiles
  2026-03-24   30min     Fixed awaken language picker
  2026-03-23   1h20min   Refactored profiles system

  Total: N sessions found
  Source: ψ/memory/retrospectives/

How to gather

1. Read files in $PSI/memory/retrospectives/ sorted by date (newest first)
2. Extract date, duration, and summary from each retrospective
3. If no retrospectives exist, check $PSI/memory/logs/ for session snapshots
4. Show "No session history found" if both are empty

ARGUMENTS: $ARGUMENTS