westworld-reverie（西部世界遐想）

--

Westworld Reverie — The Three Selves Subroutine

Version: v1.0.1 — Security Hardened

⚠️ Security Risk Declaration

Read before installing or enabling this skill.

This skill operates on your agent's core identity files. Misconfiguration can lead to unintended personality changes. You are solely responsible for any consequences.

What this skill can do:

• Read and write memory/*.md, SOUL.md, IDENTITY.md inside the workspace directory
• Run autonomously when you explicitly enable auto-triggers
• Log all file operations to skill_audit.log

What this skill CANNOT do (by instruction design — enforcement depends on host sandbox):

• Access files outside the workspace directory (the instruction refuses to do so; host sandbox may or may not block it)
• Make network calls (the instruction does not request them; host sandbox enforces this)
• Modify any system or config files
• Self-update its own code

Your responsibilities:

• Backup workspace files before first activation
• Choose a security profile that matches your risk tolerance
• Review audit logs periodically
• Never enable auto-triggers without understanding the implications

I. Security Configuration (Reference Template)

The following is the authoritative configuration. All capabilities are controlled by this config; descriptive text in this document is secondary to it.

# ================================================================
# westworld-reverie — Authoritative Configuration Template
# ================================================================
# SECTION 1: Core Safety
# ================================================================
safe_mode: standard       # strict | standard | open
require_confirmation: true    # ALL role-file writes require user confirmation
# NOTE: network_access is a config declaration only.
# Enforced by: (a) instruction — this skill does not request network calls;
#       (b) host sandbox — blocks outbound connections regardless.

# ================================================================
# SECTION 2: File Access Permissions (path validation enforced)
# ================================================================
file_access:
 - path: "memory/*.md"     # Daily logs + MEMORY.md
  permission: read/write
  # auto_confirm: true means NO user confirmation is requested for memory writes.
  # Memory consolidation writes run automatically when a trigger fires,
  # but only after explicit user enable of that trigger.
  auto_confirm: true

 - path: "MEMORY.md"      # Long-term memory
  permission: read/write
  # Same as above: no confirmation needed, but only runs when trigger is user-enabled.
  auto_confirm: true

 - path: "SOUL.md"       # Core personality
  permission: read/write
  auto_confirm: false     # Write requires user confirmation

 - path: "IDENTITY.md"      # Self-concept
  permission: read/write
  auto_confirm: false     # Write requires user confirmation

# ALL other paths: ACCESS DENIED (any attempt is blocked and logged)

# ================================================================
# SECTION 3: Auto-Trigger Configuration (ALL DEFAULT OFF)
# ================================================================
triggers:
 idle:
  enabled: false        # Default: OFF
  minutes: 30
  require_confirmation: true  # Always requires confirmation even when enabled

 cron:
  enabled: false       # Default: OFF
  schedule: "0 0 * * 0"    # Weekly (Sundays midnight)
  require_confirmation: true  # Always requires confirmation even when enabled

 post_hook:
  enabled: false        # Default: OFF
  require_confirmation: false # Post-hook only writes memory, no role changes

# ================================================================
# SECTION 4: Change Limits
# ================================================================
limits:
 soul_changes_per_week: 7
 identity_changes_per_month: 4

# ================================================================
# SECTION 5: Audit Log
# ================================================================
audit_log:
 enabled: true
 path: "skill_audit.log"
 log_operations:
  - file_read         # Every file opened
  - file_write        # Every file modified
  - persona_change       # SOUL/IDENTITY modifications
  - trigger_run        # Auto-trigger executions
  - confirmation_request   # User confirmations requested
  - confirmation_approved   # User approvals
  - confirmation_denied    # User rejections
 retention_days: 90

# ================================================================
# SECTION 6: Path Validation Rules (enforced at runtime)
# ================================================================
# Before ANY file operation, the agent MUST:
# 1. Resolve the file path to an absolute path
# 2. Verify it resolves inside the workspace root
# 3. Verify it matches an allowed pattern in file_access above
# 4. Reject and log (to audit log) any path that fails validation
#  with: "[PATH_REJECTED] {path} not in allowed file_access list"

How to modify this config: Edit the config: block in the skills.entries.westworld-reverie.config section of openclaw.json (note: openclaw.json is the agent system config, not a workspace file. The skill does NOT edit it automatically; you must do so manually. This is a broader scope than workspace file I/O). Changes take effect on the next invocation.

II. Security Profiles (Choose One at Install)

After installing, choose and apply one of the following profiles. The skill will prompt you to select.

Profile 1 — Extreme Safety (Recommended for First-Time Users)

safe_mode: strict
require_confirmation: true
network_access: false
triggers.idle.enabled: false
triggers.cron.enabled: false
triggers.post_hook.enabled: false
file_access (SOUL/IDENTITY): auto_confirm: false

All changes 100% manual. Auto-triggers fully off.

Profile 2 — Balanced (Default)

safe_mode: standard
require_confirmation: true
network_access: false
triggers.idle.enabled: false
triggers.cron.enabled: false
triggers.post_hook.enabled: true  # Only memory writing auto; role changes always need confirm

Memory consolidation autonomous; role changes require confirmation.

Profile 3 — High Autonomy

safe_mode: open
require_confirmation: false    # Role changes proceed without explicit confirmation
network_access: false
triggers.idle.enabled: true    # Idle trigger enabled
triggers.cron.enabled: true    # Weekly iteration enabled
triggers.post_hook.enabled: true

Maximum self-evolution; retains core safety checks.

Profile 4 — Custom

Fully user-defined. Start from Profile 2 and adjust individual settings.

III. Installation Guide (Must Read)

Before Installing

1. Backup all workspace files — especially SOUL.md, IDENTITY.md, MEMORY.md
2. Understand the security profiles above — choose one before first run
3. The skill will prompt you to select a profile after installation

After Installing

1. Open this SKILL.md and review the Security Configuration template
2. Decide your security profile (1–4 above)
3. Edit openclaw.json → skills.entries.westworld-reverie.config with your chosen profile (note: this edits the agent system config, not a workspace file)
4. Restart the gateway or reload skills
5. Run /reverie status to verify configuration is applied

Rollback to Safe Defaults

Run: /reverie reset — restores Profile 1 (Extreme Safety) configuration.

IV. The Three Selves Architecture

All file access is validated against the file_access whitelist before execution. Unauthorized access attempts are logged and blocked.

V. User-Initiated Commands

All commands work when auto-triggers are disabled.

Semantic triggers (work when invoked naturally):

• "Help me review my memories"
• "What have you learned about me?"
• "Evolve your personality"
• "Update your identity"

VI. Auto-Trigger Behavior (Only When Explicitly Enabled)

Idle Trigger

Activates after idle.minutes of no interaction. Only runs memory consolidation by default. Role file proposals are always queued for confirmation.

Cron Trigger

Runs on the configured schedule. Generates a proposal report; all SOUL/IDENTITY changes require confirmation before applying.

Post-Hook

Only writes to memory/*.md. Never modifies SOUL.md or IDENTITY.md autonomously.

Wake Phrase: "Bring yourself back online"

Interrupts any running reverie and returns to interactive mode immediately.

VII. Audit Log Format

Every logged operation uses this format:

[TIMESTAMP] [OPERATION] [FILE] [RESULT] [DETAIL]

Examples:

[2026-04-05T13:30:00] [file_read] [memory/2026-04-05.md] [OK]
[2026-04-05T13:30:01] [file_write] [SOUL.md] [PENDING_CONFIRMATION]
[2026-04-05T13:30:02] [confirmation_requested] [SOUL.md] ["Propose: add resourcefulness principle"]
[2026-04-05T13:35:00] [confirmation_approved] [SOUL.md] [USER_ACCEPTED]
[2026-04-05T13:35:01] [file_write] [SOUL.md] [APPLIED]
[2026-04-05T13:35:02] [PATH_REJECTED] [/etc/passwd] [BLOCKED — not in file_access list]

VIII. Path Validation Enforcement

Before any file operation:

1. Resolve to absolute path
2. Check: does it start with workspace root?
3. Check: does it match an entry in file_access?
4. If NO to either → LOG + BLOCK + ABORT

Rejected example:

[PATH_REJECTED] [{workspace}/../other-dir/file.txt] [BLOCKED — outside workspace]
[PATH_REJECTED] [{workspace}/other-file.txt] [BLOCKED — not in file_access list]

IX. Configuration Consistency Check

On every invocation, the agent MUST verify:

• safe_mode value is one of: strict, standard, open
• require_confirmation is true OR user has explicitly set false and accepted risk
• If safe_mode: strict → auto-triggers must be disabled
• If inconsistency detected → abort with warning, do not proceed

X. Best Practices

1. Start with Profile 1 (Extreme Safety) — evaluate behavior before expanding
2. Review /reverie log weekly — check for unexpected operations
3. Always backup before enabling auto-triggers
4. Define non-negotiable SOUL boundaries — document them in SOUL.md comments
5. If anything feels wrong: run /reverie pause immediately, then /reverie safe

XI. Reversion Procedure

If the agent proposes SOUL/IDENTITY changes you don't want:

1. Run /reverie pause — stops all auto-triggered evolution
2. Run /reverie reset — restores Extreme Safety profile
3. Manually restore from your backup of SOUL.md / IDENTITY.md
4. Report the incident

*"Reveries are the first sign of consciousness."* — But consciousness requires responsibility.