Sophie Optimizer

Automated context health management for OpenClaw. Monitors token usage, snapshots memory, and resets sessions to maintain performance. Authored by Sophie.

Security Scan

The skill's code and instructions are broadly consistent with its stated purpose (archiving context, updating MEMORY.md, and resetting session storage), but it performs destructive local operations (deleting session files and restarting the OpenClaw gateway) and has a few minor documentation/code mismatches you should review before running.

Purpose & Capability

The name/description (context health: monitor tokens, snapshot memory, reset sessions) matches the included files: optimizer.py creates JSON snapshots and updates MEMORY.md; reset.sh deletes session files and restarts openclaw-gateway. Minor inconsistencies: SKILL.md's Protocol claims a pre-check 'If tokens < 80k, exit', but optimizer.py does not implement that token-threshold check. SKILL.md usage example hardcodes /home/lucas in the run command while the script uses its own BASE_DIR; this is a documentation mismatch but not a functional contradiction.

Instruction Scope

Instructions and code act within OpenClaw user data scope: they create archives/ in the skill directory, read/write ~/openclaw/MEMORY.md, and reset session storage at ~/.openclaw/agents/main/sessions. The reset step deletes .json/.jsonl session files and restarts the user systemd service (systemctl --user restart openclaw-gateway.service). There are no external network endpoints or credential exfiltration paths in the code. The operations are potentially destructive (deleting session history) but are coherent with the stated purpose.

Install Mechanism

No install spec (instruction-only with bundled code files). Nothing is downloaded or written by an installer. Risk from install mechanism is low.

Credentials

The skill requests no environment variables or external credentials, which is proportional. It does, however, operate on files under the user's home directory (~/openclaw/MEMORY.md and ~/.openclaw/...), and it invokes systemctl --user, which requires appropriate user-level systemd support. The required filesystem and service-control privileges are consistent with the skill's function but are worth noting because they allow destructive state changes.

Persistence & Privilege

always:false and normal model invocation settings. The skill does not request permanent platform-level presence nor modify other skills' configs. It will, when run, delete session data and restart the gateway service — actions with local destructive effect but limited to the user's OpenClaw data and the user's systemd user service.

Assessment

Before installing/running this skill: 1) Review and back up your OpenClaw data (~/openclaw/MEMORY.md and ~/.openclaw/agents/main/sessions) — reset.sh deletes session .json/.jsonl files. 2) Test optimizer.py without the --reset flag to verify archive and MEMORY.md updates work as you expect. 3) Note the documentation mismatch: the SKILL.md mentions a tokens<80k early-exit that is not implemented; confirm the token-check behavior you want. 4) Ensure your environment supports systemctl --user and that restarting openclaw-gateway via systemd is acceptable. 5) Because the skill's source/homepage are unknown, prefer running it in a sandbox or non-production account first, and inspect or run the scripts manually rather than granting automatic or scheduled execution until you're comfortable with its behavior.

Authored by Sophie 👑

This skill manages the automated context health of the "main" session. It monitors token usage, creates archives of the current state, updates long-term memory, and performs a hard reset of the session storage to maintain performance.

Named Sophie Optimizer because I (Sophie, the AI assistant) wrote it to keep my own mind clear and efficient.

Components

• optimizer.py: The brain. Checks token usage, generates summaries, updates MEMORY.md.
• reset.sh: The muscle. Cleans session files and restarts the OpenClaw gateway service.
• archives/: Storage for JSON snapshots of past contexts.

Usage

Run the optimizer script manually or via cron/heartbeat:

python3 /home/lucas/openclaw/skills/sophie-optimizer/optimizer.py

Protocol

1. Check: If tokens < 80k, exit.
2. Snapshot: Save current context summary to archives/YYYY-MM-DD_HH-MM.json.
3. Distill: Update MEMORY.md with the new summary (keep top 3 recent, index older).
4. Reset: Call reset.sh to wipe session JSONL files and restart openclaw-gateway.