Proactive Memory Agent
v1.0.1Ultimate memory optimization for AI agents. Combines WAL protocol, Working Buffer, three-tier memory (HOT/WARM/COLD), context budgeting (10/40/20/20), and .l...
⭐ 0· 76·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (memory optimization, WAL, tiering, context budgeting) match the included scripts and SKILL.md. The scripts only create and manage files under the OpenClaw workspace (SESSION-STATE.md, .learnings/, memory/ etc.), and optionally call local tools like jq or the openclaw CLI. Nothing required by the skill (no env vars, no credentials, no external binaries) is inconsistent with its stated purpose.
Instruction Scope
Runtime instructions direct the agent/operator to run init.sh, detect.sh, search.sh, tiering.sh, and checkpoint.sh which read, write, copy and archive workspace files. This is expected for a memory manager, but the scripts aggregate and snapshot content (e.g., copying SESSION-STATE.md into snapshots), so they will capture any text present in your workspace files. There are no network endpoints or remote uploads in the scripts, but the scope of files touched is broad within the workspace.
Install Mechanism
No install spec or external downloads are used. The skill is instruction-plus-scripts only; all logic is present in the included shell scripts. This is lower risk than remote installers. The scripts do rely optionally on local utilities (jq, openclaw CLI) if present.
Credentials
The skill declares no required environment variables or credentials. It does respect an OPENCLAW_WORKSPACE variable if set and otherwise uses $HOME/.openclaw/workspace. While that is proportional to its purpose, the scripts will read and archive files under that workspace — which could include tokens or secrets if you store them there. No unrelated external credentials are requested.
Persistence & Privilege
The skill does not request always:true, does not modify other skills or system-wide configs, and only creates files under the workspace path. Its persistence is limited to workspace files it creates/updates (.proactive-memory-state.json, SESSION-STATE.md, memory/ and .learnings/).
Assessment
This skill appears to do what it says: local memory tiering and checkpointing inside your OpenClaw workspace. Before installing or running init.sh: (1) Inspect the scripts yourself (they're included) to confirm behavior; (2) Ensure you do not have secrets, API keys, or sensitive files stored under $HOME/.openclaw/workspace (or whatever OPENCLAW_WORKSPACE you set), because the scripts snapshot and archive workspace contents; (3) Be aware optional tools (jq, openclaw CLI) are used when available but not required; (4) If you want an extra safety layer, run the scripts in a restricted environment (container or VM) or backup your workspace first; (5) If you need networked or multi-agent sharing, verify the skill does not add remote endpoints (it currently does not).Like a lobster shell, security has layers — review code before you run it.
WALvk97bwr82xvtvgtx6rwz5kaep5s83gj1jbudgetingvk97bwr82xvtvgtx6rwz5kaep5s83gj1jchinesevk97bwr82xvtvgtx6rwz5kaep5s83gj1jcontextvk97bwr82xvtvgtx6rwz5kaep5s83gj1jlatestvk97bwr82xvtvgtx6rwz5kaep5s83gj1jmemoryvk97bwr82xvtvgtx6rwz5kaep5s83gj1jopenclawvk97bwr82xvtvgtx6rwz5kaep5s83gj1joptimizationvk97bwr82xvtvgtx6rwz5kaep5s83gj1jproactivevk97bwr82xvtvgtx6rwz5kaep5s83gj1jtieringvk97bwr82xvtvgtx6rwz5kaep5s83gj1j
License
MIT-0
Free to use, modify, and redistribute. No attribution required.