| name | setup-agents-memory |
| description | Configure Agents Memory for the current agent client. Use only when the user explicitly invokes /setup-agents-memory or asks to install, set up, repair, verify, or configure Agents Memory hooks for the agent currently running this skill. |
| disable-model-invocation | true |
Setup Agents Memory
Configure only the current agent client. Do not configure other agents.
Workflow
- Verify the internal tool:
~/.agents-memory/bin/agents-memory doctor
- Detect the current agent client from the runtime context and filesystem (the identifier in backticks is what you pass as
--agent):
codex (Codex): ~/.codex
claude (Claude): ~/.claude
opencode (OpenCode): ~/.config/opencode
- Ensure this agent has symlinks to all three skills. For each skill (
get-learnings, save-learnings, setup-agents-memory), check whether ~/.<agent>/skills/<skill> exists and points to ~/.agents-memory/skills/<skill>. If a symlink is missing, create it:
ln -s "$(python3 -c "import os; print(os.path.relpath('$HOME/.agents-memory/skills/<skill>', '$HOME/.<agent>/skills'))")" ~/.<agent>/skills/<skill>
Do not overwrite an existing file or symlink without explicit user confirmation.
- Invoke the internal tool to configure the current agent only:
~/.agents-memory/bin/agents-memory setup --agent <detected-agent> --cwd "$PWD"
- If the hook or plugin format for the current agent is unclear, look up current official documentation before editing config.
- Validate the setup by running:
~/.agents-memory/bin/agents-memory startup --cwd "$PWD"
- Report exactly what was configured and what was skipped.
Rules
- Do not touch project files.
- Do not configure or suggest configuration for other agent clients.
- Do not invent hook syntax; verify official docs when unsure.
- Do not create root directories for agents that are not installed.
- Do not replace conflicting files or symlinks without explicit user confirmation.