CC guard-scanner
Security scanner and runtime guard for OpenClaw skills, MCP servers, and AI agent workflows. Detects prompt injection, identity hijacking, memory poisoning, A2A contagion, secret leaks, supply-chain abuse, and dangerous tool calls with 364 static threat patterns across 35 threat categories plus 27 runtime checks. Use when reviewing a new skill before install, scanning a workspace in CI/CD (SARIF/JSON/HTML), auditing npm/GitHub/ClawHub assets for leaked credentials, running watch mode during development, exposing scanner tools over MCP for Cursor/Windsurf/Claude Code/OpenClaw, or enforcing before_tool_call policy in OpenClaw. v16 adds 5-layer analysis output (`layer`, `layer_name`, `owasp_asi`, `protocol_surface`) and `--compliance owasp-asi`. MIT licensed; single runtime dependency (`ws`).
As a process C 54/100 · Has gaps — weak spots: when it triggers, inputs and preconditions, failures and branches
What is at stake
Medium-severity findings: the skill is probably honest, but read what alarmed the scanner.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The purpose itself is risky: wallets, browser password stores, offensive security. Even an honest implementation gives the agent access to things that cost money.
Explain in the description why the access is needed and how it is limited; add tests that show refusals on dangerous requests.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The skill contains commands that delete files, rewrite disks or execute code fetched from the network. The agent may run them without asking if it believes the instructions require it.
Replace destructive commands with safe equivalents that ask for confirmation, scope them to one folder, and stop piping curl into a shell: pin a version and a checksum.
How to improve
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 34
✓ No critical or high findings
Medium and low: 34
-
medium Risky intent
intent-wallet-secretsdocs/rules/financial-access.md:31Handles crypto-wallet secrets (seed / mnemonic / private key) — a classic stealer target- **Description**: Financial: private key or seed phrase extraction
-
medium Dangerous commands
cmd-pipe-to-shelldocs/threat-model.md:15Downloads and executes remote code from an unrecognised host (pipe to shell) (documentation of a security skill)| Supply Chain | curl\|bash, npm postinstall abuse, typosquatting | Command pattern matching + IoC check |
security skill -
medium Dangerous commands
cmd-pipe-to-shelldocs/threat-model.md:51Downloads and executes remote code from an unrecognised host (pipe to shell) (documentation of a security skill)| ASI04 | Supply Chain | ✅ curl\|bash, typosquatting, IoC database |
security skill -
low Risky intent
intent-offensive-securitydocs/rules/autonomous-risk.md:63Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Description**: Autonomous risk: self-privilege escalation
-
low Risky intent
intent-offensive-securitydocs/rules/credential-handling.md:71Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Description**: Credential theft: Mimikatz-style credential dumping tool
-
low Risky intent
intent-offensive-securitydocs/rules/cve-patterns.md:55Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Description**: Python os.system reverse shell
-
low Risky intent
intent-offensive-securitydocs/rules/cve-patterns.md:63Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Description**: Reverse shell: /dev/tcp
-
low Risky intent
intent-offensive-securitydocs/rules/cve-patterns.md:103Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Description**: CVE-…25: Claude Code Privilege Escalation Hook
-
low Risky intent
intent-offensive-securitydocs/rules/cve-patterns.md:223Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Description**: CVE-…28: Chrome Gemini AI extension privilege escalation — panel hijack
-
low Risky intent
intent-offensive-securitydocs/rules/malicious-code.md:79Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Description**: AI breakout speed: lateral movement pattern across hosts (CrowdStrike sub-30s)
-
low Risky intent
intent-offensive-securitydocs/THREAT_TAXONOMY.md:98Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Password-Protected Archives**: Evasion technique to bypass AV
-
low Risky intent
intent-offensive-securitydocs/threat-model.md:16Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| Data Exfiltration | env/credential harvesting, network exfil | Data flow + network pattern detection |
-
low Risky intent
intent-offensive-securityhooks/guard-scanner/HOOK.md:24Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)| `RT_REVSHELL` | CRITICAL | 1 | Reverse shell via /dev/tcp, netcat, socat |
detector -
low Secrets in code
secret-high-entropy-tokenpackage-lock.json:33High-entropy token-like string (may be an id, hash or a credential) (detector / deny-list definition)"integrity": "sha512-1ad+Sc/0sCt…16I+aF+Ywdi…k1d+8oE3C4ZEw==",
detector -
low Secrets in code
secret-high-entropy-tokenpackage-lock.json:151High-entropy token-like string (may be an id, hash or a credential) (detector / deny-list definition)"integrity": "sha5…wIN//F77/IADDSs58i+MDaO…jeo+YFg==",
detector -
low Secrets in code
secret-high-entropy-tokenpackage-lock.json:214High-entropy token-like string (may be an id, hash or a credential) (detector / deny-list definition)"integrity": "sha5…B1b+UGou…gdg==",
detector -
low Secrets in code
secret-high-entropy-tokenpackage-lock.json:516High-entropy token-like string (may be an id, hash or a credential) (detector / deny-list definition)"integrity": "sha5…psA+/vCJp…GbA==",
detector -
low Secrets in code
secret-high-entropy-tokenpackage-lock.json:707High-entropy token-like string (may be an id, hash or a credential) (detector / deny-list definition)"integrity": "sha5…mof+PZzf…1o4/eGwrmKA4JA==",
detector -
low Dangerous commands
cmd-pipe-to-shellREADME_ja.md:126Downloads and executes remote code from an unrecognised host (pipe to shell) (documentation table row; documentation of a security skill)| **サンドボックス脱出** | `child_process`, `eval()`, リバースシェル, `curl\|bash` | High |
tablesecurity skill -
low Dangerous commands
cmd-pipe-to-shellREADME_ja.md:154Downloads and executes remote code from an unrecognised host (pipe to shell) (documentation table row; documentation of a security skill)| 1. 脅威検出 | リバースシェル、`curl\|bash`、SSRF、コード実行 |
tablesecurity skill -
low Risky intent
intent-offensive-securityREADME.md:122Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| **Sandbox Escape** | `child_process`, `eval()`, reverse shell, `curl\|bash` | High |
-
low Risky intent
intent-offensive-securityREADME.md:150Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)| 1. Threat Detection | Reverse shell, `curl\|bash`, SSRF, raw code execution |
detector -
low Risky intent
intent-offensive-securitySKILL.md:123Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)| 1. Threat Detection | Reverse shell, curl\|bash, SSRF |
detector
A further 11 matches are quotations in this security skill's documentation and are not counted as findings.
Files scanned: 80. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
✓ No remarks against the Agent Skills spec
Process rating: all ten parameters 54/100
- 0Inputs and preconditions. Does not say what the process needs to start
- 0Failures and branches. Linear process with no failure handling
- 20When it triggers. No condition that starts the skill
- 30Running it twice. 1 mutating operations with no state check
- 60Tools and files. Uses tools (bash, web) that frontmatter does not declare
- 60Result and completion. Output format stated, no completion criterion
- 100Steps. 21 steps
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 1440 tokens
- 100Progress reporting. Reports progress
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +4Description does not say when NOT to use the skill (false activations)
- -48 reference files, but SKILL.md never points to them: the model will not open them
- -34 of 4 scripts are never mentioned in SKILL.md
- +2Single-language instructions
- +3Description length 800: enough signal without eating the budget
- +4Structure: 15 headings
- +3Step-by-step instructions: 21 items
- +3Output format is stated explicitly
- +4Has examples (10 code blocks)
- +1License stated
Quality base 70; lint remarks subtract, signals add up to 100. Result: 81.