How does the Dory-Proof Memory System ensure continuity after a session reset?

It writes the user’s exact words to state/ACTIVE.md immediately, then uses a defined boot sequence to pull the latest context from state/HOLD.md, state/ACTIVE.md, state/DECISIONS.md, memory/recent-work.md, and MEMORY.md. This guarantees that the original intent is preserved and re-applied after a reset.

What happens if the user input is longer than the MEMORY.md limit?

Long-term memory is curated and kept under ~10KB. For larger contexts, you store the most critical excerpts in MEMORY.md, and break the remainder into structured, referenceable chunks in memory/ or memory/recent-work.md with proper indexing and timestamps. The goal is to preserve essential intent and decisions while avoiding drift.

How do I handle conflicts between a new instruction and a blocked item in HOLD?

HOLD entries take precedence. If a new instruction appears to conflict with an existing HOLD, the agent must resolve the conflict by updating HOLD.md with a justified reason for blocking and pausing the task until the hold is cleared or the rationale is updated.

Can I reset memory or clear a session’s history?

Yes. You can reset the active state by clearing or archiving state files, but you should never delete AGENTS.md or MEMORY.md without documenting the rationale. When in doubt, create an entry in state/DECISIONS.md for the reset decision and ensure memory.md reflects the updated long-term state.

How is time and sequencing tracked?

DECISIONS.md stores timestamps for decisions. The boot sequence uses chronological reads (HOLD, ACTIVE, DECISIONS, memory, MEMORY) so you can reconstruct the sequence of actions and intent across sessions.

What security considerations should I keep in mind?

Store memory on local disk within controlled directories. Avoid exposing sensitive data in memory.recent-work.md or MEMORY.md. Implement access controls and consider encryption for sensitive payloads if your environment requires it.

Dory-Proof Memory System

Verified

@justinhartbiz

npx machina-cli add skill @justinhartbiz/dory-memory --openclaw

Files (1)

SKILL.md

3.7 KB

Dory-Proof Memory System

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:

IMMEDIATELY write their EXACT WORDS to state/ACTIVE.md
Then interpret what it means
Then do the work
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)

Read state/HOLD.md — what's BLOCKED
Read state/ACTIVE.md — current task
Read state/DECISIONS.md — recent choices
Read memory/recent-work.md — last 48 hours
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):

state/HOLD.md — blocks override all
state/ACTIVE.md — current instruction
state/DECISIONS.md — recent choices
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

Source

git clone https://clawhub.ai/justinhartbiz/dory-memoryView on GitHub

Overview

Dory-Proof Memory System provides persistent, on-disk memory for AI agents that forget between sessions. By enforcing a disciplined, file-based workspace and the Dory-Proof pattern, you achieve continuity across context resets. This makes it ideal for setting up agent memory, structuring workspaces, tracking tasks, and preventing context-loss errors. The agent reads and writes to disk as the single source of truth, while boot-time checks ensure ongoing consistency.

How This Skill Works

Core Principle: Text > Brain. Write everything down. Files are memory. The agent only "remembers" what’s on disk. Dory-Proof Pattern (critical): When the user gives a task, you must follow these steps in order: 1) IMMEDIATELY write the user’s EXACT WORDS to state/ACTIVE.md 2) Interpret what it means 3) Do the work 4) Mark complete when finished Why this works: paraphrasing introduces drift across context flushes. Storing exact quotes preserves intent when the session context is cleared. Workspace Structure: Use a well-defined directory layout to separate active work from long-term memory and system rules. The structure is designed to minimize drift and maximize traceability. Boot Sequence (Every Session): 1) Read state/HOLD.md to surface any BLOCKED items 2) Read state/ACTIVE.md to identify the current task 3) Read state/DECISIONS.md for recent choices with timestamps 4) Read memory/recent-work.md for the last 48 hours of activity 5) Read MEMORY.md for curated long-term memory (main session only) Output a status line after boot, e.g.: 📋 Boot: ACTIVE=[task] | HOLD=[n] items | STAGING=[n] drafts State File Formats (high level): - state/ACTIVE.md holds the exact user words and the interpretation, plus a checklist of steps to completion - state/HOLD.md lists blocked items with dates and reasons - state/DECISIONS.md logs recent decisions with timestamps - memory/recent-work.md (historical) tracks recent activity - MEMORY.md stores long-term memory under ~10KB of curated content Conflict Resolution: When files conflict, priority is: HOLD (blocks) > ACTIVE (current instruction) > DECISIONS (recent choices) > AGENTS.md (general rules).

When to Use It

Set up agent memory for long-running tasks or multi-session assistants
Build and maintain workspace structure for scalable agents
Implement task tracking and decision persistence across sessions
Prevent context-loss errors during resets or timeouts
Onboard new agents that must retain knowledge across restarts

Quick Start

Copy template files from assets/templates into your workspace to establish the Dory-Proof structure: cp -r skills/dory-memory/assets/templates/* ~/.openclaw/workspace/
Customize core identity and human-facing context by editing SOUL.md and USER.md
Start a session and begin interacting. On the first user input, the Dory-Proof pattern will trigger automatically when you implement the pattern in code.

Best Practices

Always write the exact user words to state/ACTIVE.md before any interpretation or action
Keep state files minimal and avoid drift by not paraphrasing in ACTIVE.md
Check state/HOLD.md before acting on any task that looks ready
Record decisions in state/DECISIONS.md with precise timestamps to aid traceability
When memory growth approaches the 10KB limit, trim or summarize long-term memory in MEMORY.md without sacrificing critical context
Regularly refresh memory/recent-work.md to reflect the last 48 hours of activity
Keep AGENTS.md as the authoritative operating rules document and do not rename it
Use WORKSPACE-INDEX.md to locate and navigate workspace components efficiently

Example Use Cases

Onboarding a coding assistant that must remember project context across sessions, including file paths, dependencies, and prior decisions
Bug triage workflow where tasks persist between sessions and decisions are auditable through DECISIONS.md
Multi-step automations that require continuity across resets, such as build pipelines or documentation generation
Collaborative robotics or CI/CD agents where workspace structure and task history must persist for audits

Frequently Asked Questions

Add this skill to your agents