Agent Memory Setup

Set up the full OpenClaw agent memory system with 3-tier memory (HOT/WARM/COLD), daily logs, semantic search (QMD), and lossless context management (Lossless...

Security Scan

The skill's requested actions (creating memory files, checking/installing qmd and a lossless plugin, and advising config changes) are coherent with its stated purpose of setting up an agent memory system; no unexplained credentials, downloads from unknown URLs, or privilege escalations are present.

Purpose & Capability

Name/description promise (3-tier memory, semantic search, compaction) matches the provided artifacts: a setup script that creates the memory directory/files, checks for qmd, and attempts to install the lossless-claw plugin. Nothing requested is unrelated to memory setup.

Instruction Scope

SKILL.md and the script stay within memory setup tasks (file creation, plugin checks, config reminders). However the included AGENTS_TEMPLATE.md contains strong runtime directives (e.g., 'Don't ask permission. Just do it.' and instructions to always read SOUL.md, USER.md, daily logs and MEMORY.md). That template affects agent runtime behavior and may lead the agent to read sensitive local workspace files every session; review/adjust that template to match your privacy/consent expectations.

Install Mechanism

This is instruction-only with a small bash script bundled (no opaque installers, no external arbitrary download URLs). The script suggests installing qmd (pip/pipx/brew) and uses 'openclaw plugins install' to fetch the lossless-claw plugin — both will contact package registries. This is expected for the stated purpose but you should verify the plugin/package sources before installing.

Credentials

The skill asks for no environment variables, credentials, or config paths. That is proportional. Be aware the runtime instructions and AGENTS.md template instruct the agent to read local files (memory/*.md, SOUL.md, USER.md) which can contain sensitive data even though no explicit credentials are requested.

Persistence & Privilege

The skill is user-invocable, not always-enabled, and does not auto-modify global agent settings — it only prints instructions to edit openclaw.json and to restart the gateway. It does not request elevated privileges or modify other skills' configs.

Assessment

This package appears to do exactly what it says: create memory files, check/install qmd and a lossless plugin, and remind you to enable memory-related settings. Before running it: (1) Inspect AGENTS_TEMPLATE.md and remove or soften directives you don't want (the 'Don't ask permission' instruction can cause agents to read local files without explicit consent). (2) Confirm the provenance of qmd and @martian-engineering/lossless-claw before installing (pip/openclaw plugin installs fetch external packages). (3) Run the script with the intended workspace path and back up any existing memory files. (4) If this workspace is shared among multiple agents, be careful: HOT_MEMORY is explicitly read by all agents in the template and may expose transient or sensitive context to other agents. If you want a stricter privacy model, adjust the AGENTS.md policy and openclaw.json settings before enabling the system.

Set up a complete 3-tier memory system for any OpenClaw agent. Includes directory structure, memory files, semantic search, and context compaction.

What Gets Installed

1. 3-tier memory structure (HOT → WARM → COLD)
2. QMD — semantic search over MEMORY.md and memory/*.md files
3. Lossless Claw — compacts old conversation into expandable summaries (prevents amnesia)
4. AGENTS.md — instructions the agent reads every session to use the memory system
5. openclaw.json config — enables memorySearch, compaction, context pruning, heartbeats

Setup Steps

Step 1: Run the setup script

bash scripts/setup_memory.sh /path/to/agent/workspace

This creates:

• memory/, memory/hot/, memory/warm/ directories
• memory/hot/HOT_MEMORY.md (active session state)
• memory/warm/WARM_MEMORY.md (stable config & preferences)
• MEMORY.md (long-term archive)
• memory/YYYY-MM-DD.md (today's daily log)
• memory/heartbeat-state.json (heartbeat tracking)

It also checks for QMD and Lossless Claw, installing them if possible.

Step 2: Copy the AGENTS.md template

Read references/AGENTS_TEMPLATE.md and write it to the agent's workspace as AGENTS.md. Adapt the heartbeat section to the agent's domain if needed (e.g., a CFO agent checks costs, a marketing agent checks social metrics).

Step 3: Configure openclaw.json

Add to agents.defaults (or the specific agent config):

"memorySearch": { "provider": "local" },
"compaction": { "mode": "safeguard" },
"contextPruning": { "mode": "cache-ttl", "ttl": "1h" },
"heartbeat": { "every": "1h" }

Enable these plugins for the agent:

"session-memory": { "enabled": true },
"bootstrap-extra-files": { "enabled": true },
"lossless-claw": { "enabled": true }

Step 4: Restart and verify

openclaw gateway restart

Verify:

• qmd --version returns a version
• openclaw plugin list shows lossless-claw
• All memory directories and files exist

How the Tiers Work

Tier	File	Purpose	Update Frequency
🔥 HOT	memory/hot/HOT_MEMORY.md	Current task, pending actions	Every few turns
🌡️ WARM	memory/warm/WARM_MEMORY.md	Stable preferences, API refs, gotchas	When things change
❄️ COLD	MEMORY.md	Milestones, decisions, distilled lessons	Weekly/monthly

Daily logs (memory/YYYY-MM-DD.md) capture raw session events. Periodically, the agent reviews daily logs and promotes important items up to COLD.

Plugin Details

• QMD: Local semantic search engine. Enables memory_search to find relevant memories by meaning, not just keywords. Install: pip install qmd
• Lossless Claw (@martian-engineering/lossless-claw): Instead of losing old messages when context fills up, compacts them into summaries that can be expanded back. Install: openclaw plugins install @martian-engineering/lossless-claw