Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Bw Openclaw Boost
v1.2.4OpenClaw Boost enhances OpenClaw efficiency with cost tracking, memory management, compression, permission control, and task coordination tools.
⭐ 0· 59·1 current·1 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The skill's name/description match the included scripts: cost tracking, memory compaction, permission management and a local coordinator. However the SKILL.md repeatedly asserts it "does not access global ~/.openclaw" while the installer and SKILL layout install the skill into ~/.openclaw/bw-openclaw-boost — this is a borderline inconsistency (subdir under the global app directory). That installation target is plausible for an OpenClaw plugin but contradicts the stronger claim of not touching a global ~/.openclaw location.
Instruction Scope
Runtime instructions and scripts are local-only (install.sh copies files to ~/.openclaw/bw-openclaw-boost and tools run from there). Several tools call the local 'openclaw' CLI (subprocess.run(['openclaw','status'])) to read session/status lines. Parsing and persisting 'openclaw status' output (e.g., session keys/agent identifiers) into local logs/reports may capture sensitive session identifiers or metadata. The SKILL.md claims "only read-only openclaw CLI calls" which is true technically, but reading and storing status output can leak identifiers—this is scope-creep relative to a purely passive helper and should be considered when evaluating data exposure.
Install Mechanism
No remote download/install spec; the skill is instruction+local files with an install.sh that copies files into the user's home (~/.openclaw/bw-openclaw-boost) and sets execute bits. That is low risk from supply-chain perspective (no arbitrary network fetch).
Credentials
The skill requests no environment variables or external credentials, which fits its stated design. However multiple scripts parse and persist data derived from the local 'openclaw status' CLI (session keys, 'agent:' identifiers, cache percentages). While not requesting secrets, the code writes these identifiers to local logs/reports under the skill directory, which could be sensitive depending on what 'openclaw status' prints.
Persistence & Privilege
always:false and no special platform-wide privileges. The installer creates a per-skill directory under ~/.openclaw and writes config/memory/logs there; it does not attempt to modify other skills or global agent config beyond placing files in ~/.openclaw/bw-openclaw-boost. This is expected for a local plugin, though the choice of parent directory (the global ~/.openclaw) is worth noting.
What to consider before installing
This package appears to implement local cost, memory, compaction and coordination tools and does not fetch remote code or request credentials — but there are a few things to check before installing:
- Data exposure: several scripts run the 'openclaw status' CLI and parse its output, then write session identifiers and other status lines to local logs/reports under ~/.openclaw/bw-openclaw-boost/memory/logs and cost reports. Inspect what 'openclaw status' prints on your system; if it includes session tokens, channel keys, or other sensitive identifiers you may not want those persisted.
- Installation path vs. claims: SKILL.md says it "does not access global ~/.openclaw" but install.sh places the skill under ~/.openclaw/bw-openclaw-boost. That is a subdirectory of the app's home; this is probably fine, but if you expected the skill to live elsewhere (e.g., completely outside ~/.openclaw), be aware of that mismatch.
- Review permission_manager.py and tools/check_permission.sh: these implement local permission checks. Before enabling automation, read those files to verify the ask/allow/deny behavior and whether any actions could be escalated.
- Run offline / sandboxed first: because the tool writes local logs and can persist parsed status output, consider installing to a temporary location or running the scripts manually (python3 tools/*.py) to observe what it logs before enabling any automation.
- Minor bug: launch.sh contains a typo in the all-status branch ($PYOLS_DIR/token_budget.py). This is a non-security bug but will break that command.
If you want, I can: (1) list every file that writes to disk and where it writes; (2) scan the code for exact places that persist 'openclaw status' strings into files; or (3) produce a short checklist you can run to verify 'openclaw status' output details on your system.Like a lobster shell, security has layers — review code before you run it.
boostvk970af90bvd4exw4ye7xt3z9t5844zqdefficiencyvk970af90bvd4exw4ye7xt3z9t5844zqdlatestvk97bwbjdkykqmm9ma0hdst2mex845yd2openclawvk970af90bvd4exw4ye7xt3z9t5844zqdproductivityvk970af90bvd4exw4ye7xt3z9t5844zqdtoolsvk970af90bvd4exw4ye7xt3z9t5844zqd
License
MIT-0
Free to use, modify, and redistribute. No attribution required.