Dory-Proof Memory System

File-based memory system for AI agents that forget between sessions. Implements the "Dory-Proof" pattern for continuity across context resets. Use when setting up agent memory, building workspace structure, implementing task tracking, or preventing context-loss errors. Triggers on "memory system", "remember between sessions", "Dory pattern", "agent continuity", or "workspace setup".

Security Scan

The skill is internally coherent for a file-backed agent memory system — no external network calls, installs, or unrelated credentials — but it intentionally writes verbatim user input to disk, so you should be aware of privacy and data-retention risks before enabling it.

Purpose & Capability

The skill's name/description (file-based memory, 'Dory-Proof' pattern) matches what the instructions do: create/read/write workspace state and memory files. There are no unrelated binaries, environment variables, or external services requested.

Instruction Scope

Instructions direct the agent to read and write files under ~/.openclaw/workspace (state/, memory/, ops/, etc.) and to immediately write the user's exact words to state/ACTIVE.md before interpreting. This is coherent with the stated purpose, but it means any user-provided secret or personal data sent as a task will be persisted verbatim. There are no network endpoints or exfiltration steps in the instructions, but the guidance does not include redaction or secrets-handling rules.

Install Mechanism

No install spec or code files are present; this is instruction-only with bundled templates. Lowest-risk install posture — nothing is downloaded or executed outside the agent's normal file I/O.

Credentials

The skill requests no environment variables, no credentials, and no config paths outside the designated workspace. The requested file access is proportionate to a file-based memory system.

Persistence & Privilege

The skill instructs agents to create and persist files in the user's ~/.openclaw/workspace, which grants local persistence (expected for a memory system). It does not set always:true and does not modify other skills' configs. Consider file-permission and retention policies because data remains on disk between sessions.

Assessment

This skill appears to do what it says: implement a persistent, file-based memory for agents. Before installing, consider these points: - Privacy: the Dory pattern explicitly tells the agent to record the user's exact words verbatim. Do NOT ask the agent to 'remember' passwords, API keys, payment details, or other secrets — those will be stored on disk. - Storage location: the templates and runtime files live under ~/.openclaw/workspace; confirm you are comfortable with persistent files there and set strict filesystem permissions (e.g., restrict to your user) if needed. - Retention and pruning: follow the guidance to keep MEMORY.md small and archive/prune logs regularly to limit long-term exposure of sensitive content. - Testing: try this in an isolated or sandbox workspace first to verify the file layout and behavior. The cp command in the docs (cp -r skills/dory-memory/assets/templates/* ~/.openclaw/workspace/) assumes the asset path exists; you may need to copy templates manually if the runtime packaging differs. - Operational safeguards: if you need the agent to remember anything sensitive, use a secure secret store rather than these plaintext files; add redaction or explicit 'do not persist' rules in AGENTS.md or LESSONS.md. If you accept the tradeoffs (persistent plaintext memory vs improved continuity), the skill is consistent and can be used. If you cannot allow verbatim persistence of user content, do not enable it or modify the instructions to redact sensitive fields first.

AI agents forget everything between sessions. This skill implements a file-based memory system that survives context resets.

Core Principle

Text > Brain. Write everything down. Files are memory. The agent only "remembers" what's on disk.

The Dory-Proof Pattern (Critical)

When the user gives a task:

1. IMMEDIATELY write their EXACT WORDS to state/ACTIVE.md
2. Then interpret what it means
3. Then do the work
4. Mark complete when done

Why: Paraphrasing introduces drift. Exact words preserve intent across context flushes.

Workspace Structure

workspace/
├── AGENTS.md        # Operating rules (system file, don't rename)
├── SOUL.md          # Identity + personality
├── USER.md          # About the human
├── MEMORY.md        # Curated long-term memory (<10KB)
├── LESSONS.md       # "Never again" safety rules
├── TOOLS.md         # Tool-specific notes
│
├── state/           # Active state (check every session)
│   ├── ACTIVE.md    # Current task (exact user words)
│   ├── HOLD.md      # Blocked items (check before acting!)
│   ├── STAGING.md   # Drafts awaiting approval
│   └── DECISIONS.md # Recent choices with timestamps
│
├── memory/          # Historical
│   ├── YYYY-MM-DD.md
│   ├── recent-work.md
│   └── archive/
│
└── ops/             # Operational
    └── WORKSPACE-INDEX.md

Boot Sequence (Every Session)

1. Read state/HOLD.md — what's BLOCKED
2. Read state/ACTIVE.md — current task
3. Read state/DECISIONS.md — recent choices
4. Read memory/recent-work.md — last 48 hours
5. Read MEMORY.md — long-term (main session only)

Output status line after boot:

📋 Boot: ACTIVE=[task] | HOLD=[n] items | STAGING=[n] drafts

State File Formats

state/ACTIVE.md

## Current Instruction
**User said:** "[exact quote]"
**Interpretation:** [what you think it means]
**Status:**
- [ ] Step 1
- [ ] Step 2

state/HOLD.md

[YYYY-MM-DD HH:MM | session] Item — reason blocked

ALL agents must check before acting on anything that looks ready.

state/DECISIONS.md

[YYYY-MM-DD HH:MM | session] Decision made

Conflict Resolution

When files conflict, priority (highest first):

1. state/HOLD.md — blocks override all
2. state/ACTIVE.md — current instruction
3. state/DECISIONS.md — recent choices
4. AGENTS.md — general rules

Memory Scoring (Before Saving to MEMORY.md)

Score on 4 axes (0–3 each):

Axis	0	1	2	3
Longevity	Gone tomorrow	Weeks	Months	Years+
Reuse	One-off	Occasional	Frequent	Every session
Impact	Trivial	Nice to know	Changes outputs	Changes decisions
Uniqueness	Obvious	Slightly helpful	Hard to rederive	Impossible without

Save if: Total ≥ 8, OR any axis = 3 AND total ≥ 6.

Quick Setup

Copy template files from assets/templates/ to your workspace:

cp -r skills/dory-memory/assets/templates/* ~/.openclaw/workspace/

Then customize SOUL.md and USER.md for your agent.

References

• references/IMPLEMENTATION-GUIDE.md — Full setup walkthrough
• references/ANTI-PATTERNS.md — Common mistakes to avoid