clawbrowser
Verdict safe, but 2 findings (2 medium), mainly unauthorized tool use, resource abuse. Most recent scan 2026-02-05.
The clawbrowser skill is a legitimate browser automation tool using Playwright CLI with MODERATE security concerns. The primary issues are: (1) overly permissive wildcard tool restrictions that could be tightened, (2) missing metadata reducing transparency, (3) incomplete documentation with truncated instructions and missing referenced files, and (4) lack of resource management guidance for persistent sessions. No evidence of malicious intent (data exfiltration, prompt injection, credential theft) was found. The skill performs its stated function of browser automation but would benefit from stricter tool permissions, complete documentation, and resource management guidelines.
Scanner's overall assessment (LLM analyzer)
Second scanner: NVIDIA SkillSpector
SkillSpector found no issues but could not fully inspect the package, usually because SKILL.md refers to a file that is not present, so it fails closed to CAUTION instead of SAFE. Read it as "not fully checked", not as a finding. How the two scanners compare across the whole corpus: Two scanners, one corpus.
What this skill does
Use when the agent needs to drive a browser through the Microsoft Playwright CLI (`playwright-cli`) for navigation, form interactions, screenshots, recordings, data extraction, session management, or debugging without loading a full MCP browser. It trains the agent on the CLI commands, snapshots, and session/config habits that make Playwright CLI reliable for scripted browsing.
- Package
- 1 file, 5.4 KB
- SKILL.md
- 657 words, 8 sections
- Code
- None. Instructions only.
- Allowed tools
Bash(playwright-cli:*)
SKILL.md covers: Clawbrowser – browser control via Playwright CLI · Setup & orientation · Core interaction loop · Sessions & persistence · Tabs, navigation, and devtools · Recording, tracing, and exports · Config, state, and housekeeping · Troubleshooting & reminders
Findings
MEDIUM Overly Permissive Tool Restriction Pattern Unauthorized Tool Usellm
The allowed-tools field uses a wildcard pattern 'Bash(playwright-cli:*)' which permits execution of ANY command starting with 'playwright-cli'. This broad permission could be exploited if an attacker can influence command arguments or if the playwright-cli tool itself has vulnerabilities. The wildcard pattern defeats the purpose of granular tool restrictions.
SKILL.md
allowed-tools: Bash(playwright-cli:*)MEDIUM Potential for Unbounded Browser Sessions Resource Abusellm
The skill encourages persistent browser sessions with cookies and storage but provides no guidance on session cleanup, timeout limits, or resource management. The instructions mention 'session-stop-all' and 'session-delete' but don't enforce or recommend regular cleanup. Accumulated sessions could lead to resource exhaustion.
SKILL.md
Use --session=<name> to keep cookies, storage, and tabs isolated per workflow. Sessions behave like persistent profiles: they remember auth state, history, and tabs between commands.What is in the package
| File | Kind | Size |
|---|---|---|
SKILL.md | doc | 5.4 KB |
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/clawbrowser
- Raw scan result: clawbrowser-scan.json
- Badge for your README:
- Request a re-scan