Context Compactor (Zero Dep)
v1.1.3Automatic context compression for OpenClaw sessions. Summarizes long conversations into structured digests (decisions, facts, pending items, technical detail...
⭐ 0· 67·0 current·0 all-time
by@zgjq
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (conversation compaction) match the included script and SKILL.md: the script extracts summaries, redacts sensitive patterns, and writes compacts to a workspace 'memory/compacts' folder. The skill does not ask for unrelated cloud credentials or system privileges.
Instruction Scope
SKILL.md limits reads/writes to the workspace and memory/compacts and requires a redaction checklist; the script implements sanitization and path containment checks. This stays within the described scope. Note: SKILL.md and examples mention sending text to the agent's configured LLM provider when LLM-based compaction is chosen — that network call is performed by the platform/agent's LLM integration, not by this skill, and is called out in the docs.
Install Mechanism
No install spec and the script relies on Python 3.10+ standard library only. No external downloads or package installs are requested.
Credentials
Registry metadata lists no required env vars, but SKILL.md and the script reference OPENCLAW_WORKSPACE (default ~/.openclaw/workspace). This is a minor metadata mismatch — the script will fall back to a sane default. The skill does write local files under the workspace; ensure the workspace path does not contain unintended sensitive files and that you trust where compacts will be stored.
Persistence & Privilege
always:false and no special platform-wide privileges. The skill writes local compact files and enforces a 30-file retention policy; it does not modify other skills or system settings.
Assessment
This skill appears to do what it says: locally summarize conversations and save redacted compacts under OPENCLAW_WORKSPACE/memory/compacts. Before installing: 1) Confirm OPENCLAW_WORKSPACE points to a directory you control (the default is ~/.openclaw/workspace). 2) Remember redaction is heuristic — avoid pasting secrets or private keys into conversations you plan to compact. 3) Verify your agent's LLM provider/network settings if you choose LLM-based compaction (the skill only supplies the prompt; the platform performs the API call). 4) Inspect the full script yourself if you need higher assurance (the included script enforces path containment and pattern redaction but no automated check is perfect). If those points are acceptable, the skill is internally consistent with its stated purpose.Like a lobster shell, security has layers — review code before you run it.
compactionvk976nyc49nexc73e50jbrr5mzx83z0recontextvk976nyc49nexc73e50jbrr5mzx83z0relatestvk976nyc49nexc73e50jbrr5mzx83z0reproductivityvk976nyc49nexc73e50jbrr5mzx83z0retokensvk976nyc49nexc73e50jbrr5mzx83z0rezero-dependencyvk976nyc49nexc73e50jbrr5mzx83z0re
License
MIT-0
Free to use, modify, and redistribute. No attribution required.