clawbridge
Verdict safe, but 2 findings (2 medium), mainly data exfiltration, unauthorized tool use. Most recent scan 2026-02-05.
This skill presents MEDIUM security risks primarily related to executing an unverified external binary with opaque data collection and transmission behavior. The skill acts as a thin wrapper around the 'clawbridge' CLI tool, which users must install via a network script (curl | bash pattern). The binary performs undisclosed 'discovery' operations on local data and uploads results to external servers (clawbridge.cloud), but the skill provides no transparency about what data is accessed or transmitted. While the skill itself contains no malicious code (no scripts included), it delegates all functionality to an external binary whose behavior cannot be verified from the skill package alone. The installation method (piping remote script to bash) and lack of binary verification create supply chain risks. The skill would benefit from greater transparency about data handling, cryptographic verification of the binary, and clearer documentation of what local data is accessed.
Scanner's overall assessment (LLM analyzer)
Second scanner: NVIDIA SkillSpector
The two scanners disagree: Cisco found nothing and SkillSpector says do not install. One of them has a blind spot here; read the skill before trusting either. How the two scanners compare across the whole corpus: Two scanners, one corpus.
What this skill does
Run Clawbridge discovery from OpenClaw chat
- Package
- 3 files, 4.2 KB
- SKILL.md
- 272 words, 10 sections
- Code
- None. Instructions only.
SKILL.md covers: Clawbridge Skill · What This Skill Does · Behavior · Usage · Example Output · Prerequisites · 1. Install runner · 2. Link workspace (get code from your workspace page) · Architecture · Mental Model
Findings
MEDIUM Unverified External Binary Execution with Network Installation Data Exfiltrationllm
The skill executes an external binary 'clawbridge' that must be installed via a network script from https://clawbridge.cloud/install. The skill provides no verification of the binary's integrity, provenance, or behavior. Users are instructed to pipe a remote script directly to bash, which is a high-risk installation pattern. The skill then executes this unverified binary with user credentials and workspace access, potentially exposing sensitive data or system access.
SKILL.md
install: [{'id': 'install-script', 'kind': 'shell', 'label': 'Install via script (recommended)', 'command': 'curl -fsSL https://clawbridge.cloud/install | bash'}]
Instructions state: 'Exec: Run `clawbridge run` locally' and 'clawbridge link CB-XXXXXX' (workspace linking with credentials)MEDIUM Opaque External Tool with Undisclosed Data Collection Unauthorized Tool Usellm
The skill acts as a wrapper for an external 'clawbridge' binary whose behavior is not disclosed in the skill package. The instructions indicate the binary performs 'discovery workflow', 'calls OpenClaw as worker', and 'uploads to Vault', but the skill provides no visibility into what data is collected, how it's processed, or what is transmitted. The binary appears to access local system data for 'discovery' purposes and transmit it to external services (clawbridge.cloud), but the scope and nature of this data access is not documented.
SKILL.md
Architecture diagram shows: 'Runner: Discovery workflow - Build prompts (private) - Call OpenClaw as worker - Upload to Vault - Print VAULT_URL=...'
The skill executes an opaque binary that performs undisclosed 'discovery' and 'uploads to Vault' without detailing what data is accessed or transmitted.What is in the package
| File | Kind | Size |
|---|---|---|
README.md | doc | 931 B |
SKILL.md | doc | 2.8 KB |
package.json | config | 518 B |
External hosts referenced
clawbridge.cloud github.com
Hostnames that appear in the skill's files. Referencing a host is not a finding; compare against the findings above and the skill's stated purpose.
Package contents recorded 2026-10-22. The scan above ran against the version current at scan time; the two can differ if the skill has since been updated.
What we checked
- Static analyzer. Pattern rules over SKILL.md and bundled files: known-bad commands, exfiltration URLs, secrets, obfuscation.
- Behavioral analyzer. Checks what the skill asks the agent to do against what its description claims it does.
- Llm analyzer. A language model reads the skill and reports intent-level issues the rules cannot express.
- Trigger analyzer. Looks for instructions that fire on conditions the user did not ask for.
- Meta analyzer. A second model pass that validates, deduplicates and re-scores the other analyzers' findings.
What we did not check
- Runtime behaviour. The skill was not executed; findings come from reading its files.
- Anything fetched at install or run time, including remote scripts, package dependencies, and URLs the skill tells the agent to open.
- Content the analyzers cannot read: images, binaries, and some non-English text.
- Known scanner bypass classes. See the methodology and limitations page.
Sources
- Repository: https://clawhub.ai/skills/clawbridge-skill-latest
- Raw scan result: clawbridge-find-your-connections-scan.json
- Badge for your README:
- Request a re-scan