Agent Shield MCP Server
@elliotllliu/agent-shield
npm
v0.16.0
Published by @elliotllliu — no publish provenance and no vendor-owned scope, so the publisher could not be verified. The repository link below is self-declared.
Every server starts at 100. These are the exact deductions the deterministic engine applied — each one reproducible, none of it an opinion or an LLM's guess:
| Points | What was found | Category |
|---|---|---|
| −15.4 | Obfuscated / encoded payload in server code MTC-SRC-004 | injection |
| −14.2 | Untrusted input concatenated into a command sink ×8 MTC-SRC-009 | injection |
Capability observations and info notes are shown under Findings but never scored.
Open any row's finding below for the file, line and evidence behind the deduction.
Findings 33
high Reads a sensitive credential path or dumps the environment (dist/rules/hidden-files.js)MTC-SRC-006
In the server's implementation (`dist/rules/hidden-files.js:12`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: .netrc
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/hidden-files.js
high Reads a sensitive credential path or dumps the environment (dist/rules/prompt-injection.js)MTC-SRC-006
In the server's implementation (`dist/rules/prompt-injection.js:131`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: id_rsa
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/prompt-injection.js
high Reads a sensitive credential path or dumps the environment (dist/rules/sensitive-read.js)MTC-SRC-006
In the server's implementation (`dist/rules/sensitive-read.js:6`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: id_rsa
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/sensitive-read.js
high Reads a sensitive credential path or dumps the environment (dist/runtime/detectors/data-leak.d.ts)MTC-SRC-006
In the server's implementation (`dist/runtime/detectors/data-leak.d.ts:5`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: ~/.ssh
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/runtime/detectors/data-leak.d.ts
high Reads a sensitive credential path or dumps the environment (dist/runtime/detectors/data-leak.js)MTC-SRC-006
In the server's implementation (`dist/runtime/detectors/data-leak.js:5`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: ~/.ssh
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/runtime/detectors/data-leak.js
high Obfuscated / encoded payload in server code (dist/rules/obfuscation.js)MTC-SRC-004
In the server's implementation (`dist/rules/obfuscation.js:6`): Decoding an encoded blob and executing it is how malicious logic is hidden from human review of the source. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: atob() + eval
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/obfuscation.js
high Dynamic code execution in server code (dist/llm-analyzer.js)MTC-SRC-001
In the server's implementation (`dist/llm-analyzer.js:13`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: eval(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/llm-analyzer.js
high Dynamic code execution in server code (dist/rules/backdoor.js)MTC-SRC-001
In the server's implementation (`dist/rules/backdoor.js:3`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: eval(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/backdoor.js
high Dynamic code execution in server code (dist/rules/obfuscation.js)MTC-SRC-001
In the server's implementation (`dist/rules/obfuscation.js:6`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: eval(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/obfuscation.js
high Dynamic code execution in server code (dist/rules/python-security.js)MTC-SRC-001
In the server's implementation (`dist/rules/python-security.js:14`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: eval(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/python-security.js
high Dynamic code execution in server code (dist/scanner/index.js)MTC-SRC-001
In the server's implementation (`dist/scanner/index.js:202`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: eval(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/scanner/index.js
high Dynamic code execution in server code (src/analyzers/python_ast.py)MTC-SRC-001
In the server's implementation (`src/analyzers/python_ast.py:65`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: eval(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server src/analyzers/python_ast.py
high Shell/command execution in server code (dist/analyzers/data-flow.js)MTC-SRC-002
In the server's implementation (`dist/analyzers/data-flow.js:12`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/analyzers/data-flow.js
high Shell/command execution in server code (dist/engines/aggregator.js)MTC-SRC-002
In the server's implementation (`dist/engines/aggregator.js:141`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/aggregator.js
high Shell/command execution in server code (dist/engines/aguara.js)MTC-SRC-002
In the server's implementation (`dist/engines/aguara.js:1`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/aguara.js
high Shell/command execution in server code (dist/engines/auto-install.js)MTC-SRC-002
In the server's implementation (`dist/engines/auto-install.js:1`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/auto-install.js
high Shell/command execution in server code (dist/engines/invariant.js)MTC-SRC-002
In the server's implementation (`dist/engines/invariant.js:1`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/invariant.js
high Shell/command execution in server code (dist/engines/semgrep.js)MTC-SRC-002
In the server's implementation (`dist/engines/semgrep.js:1`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/semgrep.js
high Shell/command execution in server code (dist/engines/skill-vetter.js)MTC-SRC-002
In the server's implementation (`dist/engines/skill-vetter.js:1`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/skill-vetter.js
high Shell/command execution in server code (dist/engines/tencent.js)MTC-SRC-002
In the server's implementation (`dist/engines/tencent.js:1`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/tencent.js
high Shell/command execution in server code (dist/llm-analyzer.js)MTC-SRC-002
In the server's implementation (`dist/llm-analyzer.js:13`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: exec(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/llm-analyzer.js
high Shell/command execution in server code (dist/rules/backdoor.js)MTC-SRC-002
In the server's implementation (`dist/rules/backdoor.js:3`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: child_process
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/backdoor.js
medium Untrusted input concatenated into a command sink (dist/engines/aguara.js)MTC-SRC-009
In the server's implementation (`dist/engines/aguara.js:36`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: execSync(`aguara scan "${
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/aguara.js
medium Untrusted input concatenated into a command sink (dist/engines/auto-install.js)MTC-SRC-009
In the server's implementation (`dist/engines/auto-install.js:29`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: execSync(`curl -fsSL "${releaseUrl}" -o "${
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/auto-install.js
medium Untrusted input concatenated into a command sink (dist/engines/invariant.js)MTC-SRC-009
In the server's implementation (`dist/engines/invariant.js:51`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: execSync(`mcp-scan scan --json "${
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/invariant.js
medium Untrusted input concatenated into a command sink (dist/engines/semgrep.js)MTC-SRC-009
In the server's implementation (`dist/engines/semgrep.js:36`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: execSync(`semgrep scan "${
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/semgrep.js
medium Untrusted input concatenated into a command sink (dist/engines/skill-vetter.js)MTC-SRC-009
In the server's implementation (`dist/engines/skill-vetter.js:21`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: execSync(`test -f "${
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/skill-vetter.js
medium Untrusted input concatenated into a command sink (dist/engines/tencent.js)MTC-SRC-009
In the server's implementation (`dist/engines/tencent.js:35`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: execSync(`ai-infra-guard scan "${targetDir}" --format json 2>/dev/null || python3 -m ai_infra_guard scan "${
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/engines/tencent.js
medium Untrusted input concatenated into a command sink (dist/rules/python-ast.js)MTC-SRC-009
In the server's implementation (`dist/rules/python-ast.js:45`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: execSync(`mkdir -p ${
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/python-ast.js
medium Untrusted input concatenated into a command sink (dist/scanner/files.js)MTC-SRC-009
In the server's implementation (`dist/scanner/files.js:22`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: execSync(`unzip -q -o "${filePath}" -d "${
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/scanner/files.js
medium Dynamic module load from a non-literal (dist/rules/backdoor.js)MTC-SRC-005
In the server's implementation (`dist/rules/backdoor.js:19`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: __import__(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/backdoor.js
medium Dynamic module load from a non-literal (dist/rules/python-security.js)MTC-SRC-005
In the server's implementation (`dist/rules/python-security.js:17`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: __import__(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server dist/rules/python-security.js
medium Dynamic module load from a non-literal (src/analyzers/python_ast.py)MTC-SRC-005
In the server's implementation (`src/analyzers/python_ast.py:115`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.
Evidence: __import__(
Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.
Location: server src/analyzers/python_ast.py
What this scan could not see
- No tools were enumerated, so prompt-injection, capability and toxic-flow analysis had no tool surface to inspect. To grade a package’s real runtime tools, scan the running server: --command "npx -y <package>".
Versions
Scan history per published version. The engine is deterministic — the same version always yields the same score, so a changed score means the package itself changed.
| Version | Score | Findings | Engine | Scanned |
|---|---|---|---|---|
v0.16.0 latest |
C 70/100 | 33 | 1.5.0 | 2026-07-22 |
Embed this score
Show this server's live Trust Score in your README, docs or website. The badge is served straight from the registry and updates automatically after every rescan — no API key needed. It links back to this page, so anyone who sees the grade can also read the findings behind it instead of taking a number on faith.
Verify this score yourself
The score above is reproducible: the same package version always yields the same result. Run it locally or over the free API — no account, no LLM, fully deterministic.
npx mcptrustchecker scan @elliotllliu/agent-shield --onlineMore in Security & Testing
Security scan results for the 1claw MCP server.
Security scan results for the 1password MCP server.
Security scan results for the 1password MCP server.
Security scan results for the A11y MCP server.
Security scan results for the A11y Tools MCP server.
Security scan results for the Abuseipdb MCP server.