Smart Context Manager
v1.0.3AI-powered context management for OpenClaw sessions
⭐ 3· 3.9k·38 current·39 all-time
byPedro Gonzalez@plgonzalezrx8
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (context management) match the included script and SKILL.md. The script requires and uses the OpenClaw CLI, jq, and access to the OpenClaw state directory (~/.openclaw or OPENCLAW_STATE_DIR) — all appropriate for reading, backing up, resetting, and reinjecting session context.
Instruction Scope
Instructions and the script perform exactly what they claim: request a summary from the agent, back up the JSONL session, delete the session file to reset, and inject the summary. This is potentially destructive when run with --replace; the skill documents backups and the behavior, but users should be aware that full session contents are read, written to memory/compressed/, and (on replace) the original JSONL is deleted.
Install Mechanism
No install spec; it's an instruction-only skill with a shell script. Nothing is downloaded from external URLs or installed automatically by the skill itself.
Credentials
The skill requests no environment variables or credentials. It optionally uses OPENCLAW_STATE_DIR to locate session files (falls back to ~/.openclaw) — this is proportional and required for its function.
Persistence & Privilege
always is false and the skill does not request permanent platform-wide privileges. It modifies session files under the OpenClaw state directory (its intended scope) and writes backups into its own memory/compressed directory.
Assessment
This skill is internally consistent with its stated purpose, but be aware of these practical risks before installing: 1) The summarize+--replace operation is destructive: it deletes the session JSONL after backing it up — test with the read-only summarize command first. 2) Backups and generated summaries are stored in the skill's memory/compressed/ folder and may contain sensitive conversation data; check file permissions and retention. 3) Summarization is performed by your agent (and thus its configured model/provider), so session contents will be sent to whatever LLM backend your agent uses — ensure that is acceptable for any sensitive data. 4) The script requires the openclaw CLI and jq and read/write access to your OpenClaw state directory (~/.openclaw); review the script (compress.sh) yourself before running and keep a manual copy of important sessions if you need stronger guarantees.Like a lobster shell, security has layers — review code before you run it.
latestvk97ep0k7zztxvsvxzjfzhqccg180agktlatest-betavk976rmyjbvf0d2wkr6yyhc1fx9800p2h
License
MIT-0
Free to use, modify, and redistribute. No attribution required.