How to use the agentmemory Lifecycle Hooks agent skill

What an agent skill is and what this one does

An agent skill is a folder with a SKILL.md instruction file that an AI agent loads when a task matches it. This skill is about the agentmemory plugin's hooks. Once the plugin is installed and its server is running, routine work is remembered without the agent calling a save command: you do not have to call memory_save for normal work.

The hooks cover session start and end, which frame each unit of work and let handoff resume it; tool-use hooks, which capture what changed and why; prompt-submit, which captures intent; pre-compact, which preserves context before the host trims it; and a post-commit hook that links commits to sessions.

Install the agentmemory plugin for your host

For Claude Code, add the marketplace and install the plugin with the commands below. For Devin CLI and Devin Desktop, install the plugin and start a new session, then run /hooks inside Devin to confirm the hooks are loaded. Plugin hooks run only in local Devin sessions, not cloud sessions.

Pi and OMP use agentmemory connect commands with --with-mcp, DSH uses connect dsh --with-hooks, and Cursor's lifecycle hooks come from its enabled plugin. Hermes and OpenClaw follow their own integration setup instructions. Installing shared skills or connecting MCP alone does not enable automatic capture.

Install the skill from the ZIP or the Claude apps

Download the ZIP for this skill from this page and unzip it into ~/.claude/skills/ for all your projects, or into a project's .claude/skills/ folder for one project only. In the Claude apps, you can instead upload the skill under Settings, then Capabilities, then Skills.

The skill folder contains SKILL.md, REFERENCE.md and TROUBLESHOOTING.md. Keep them together so the links between files work.

• Personal: ~/.claude/skills/
• Single project: .claude/skills/ inside the project
• Claude apps: Settings, then Capabilities, then Skills

Trigger the skill with natural requests

The skill is set to user-invocable false, so you do not call it as a slash command. Instead, ask in plain language about the situation it covers. The skill description says it applies when explaining how memory gets captured without manual saves, when debugging missing observations, or when tuning what gets recorded.

Good requests name the symptom or the goal, and mention the host you use. The agent can then load the hook guidance and check your setup.

• Explain how agentmemory captures my Claude Code session without manual saves.
• My agentmemory observations are missing from yesterday's Devin session. Help me find out why.
• Which lifecycle events does agentmemory record, and how can I stop it recording prompts?
• Check whether the agentmemory hooks are loaded in my Devin CLI session.

Check that capture works and debug missing observations

Missing observations usually come from one of three things: the plugin is not enabled, the agentmemory server is not running, or the host is not loading the hooks. Confirm each one first. Open the viewer at http://localhost:3113 when using the default port and look for recorded observations after a real turn. A pending capture queue is not proof that the daemon stored the event.

If the problem persists, the TROUBLESHOOTING.md file in the skill covers missing tools, MCP connections and host-specific checks. In Claude Code, inspect /plugin list and /mcp. Restart or reload the host connection after configuration changes.

• Confirm the plugin is enabled in your host.
• Confirm the local agentmemory server is running.
• Run one real prompt and check the viewer at localhost:3113.
• Do not run another host's setup commands.

Tune what gets recorded

The plugin hooks capture by default without an LLM. Two separate options change behaviour: AGENTMEMORY_AUTO_COMPRESS, which compresses each observation with an LLM, and AGENTMEMORY_INJECT_CONTEXT, which injects context through hooks. Session summaries can use a configured LLM provider independently of automatic compression.

Native integrations have their own recall and capture settings, so inspect your host configuration rather than assuming these flags control every plugin. The agentmemory-config skill covers the capture and injection flags in more detail.