Anvilterm — the complete, unedited output of the deterministic mcptrustchecker engine v1.13.0, scanned . Every finding, capability tag and score component below is exactly what the engine produced — no AI, no post-processing.
{
"tool": {
"name": "mcptrustchecker",
"version": "1.13.0",
"methodologyVersion": "mcptrustchecker-1.13"
},
"target": {
"id": "anvilterm",
"source": {
"kind": "package",
"origin": "anvilterm"
},
"server": {
"name": "anvilterm"
}
},
"grade": "C",
"score": {
"score": 77,
"threatScore": 89,
"grade": "C",
"band": "C",
"categorySubtotals": {
"injection": 9.45,
"exfiltration": 0,
"permissions": 0,
"supply-chain": 1.2,
"network": 0,
"hygiene": 0
},
"vector": [
{
"kind": "threat",
"ruleId": "MTC-SRC-009",
"category": "injection",
"severity": "medium",
"confidence": "strong",
"rawWeight": 9,
"confidenceMult": 0.7,
"diminishingFactor": 1,
"appliedPenalty": 6.3
},
{
"kind": "threat",
"ruleId": "MTC-SRC-009",
"category": "injection",
"severity": "medium",
"confidence": "strong",
"rawWeight": 9,
"confidenceMult": 0.7,
"diminishingFactor": 0.5,
"appliedPenalty": 3.15
},
{
"kind": "threat",
"ruleId": "MTC-SUP-010",
"category": "supply-chain",
"severity": "low",
"confidence": "heuristic",
"rawWeight": 3,
"confidenceMult": 0.4,
"diminishingFactor": 1,
"appliedPenalty": 1.2
},
{
"kind": "client",
"term": "capability-exposure",
"level": "critical",
"label": "capability blast radius (critical) — client exposure if the model is manipulated",
"appliedPenalty": 10
},
{
"kind": "client",
"term": "verification-discount",
"level": "none",
"label": "publisher verification (unlinked) — no provenance/repo link, but the shipped source was fully read",
"appliedPenalty": 2
},
{
"kind": "client",
"term": "coverage-honesty",
"level": "source",
"label": "inspection depth (source) — how much of the target the scan could see",
"appliedPenalty": 0
}
],
"gatesFired": [],
"methodologyVersion": "mcptrustchecker-1.13"
},
"capability": {
"level": "critical",
"reasons": [
"ingests untrusted external content (a prompt-injection entry point)",
"can send data / act on an external service",
"reads sensitive or local data",
"has a read → egress path (a data-exfiltration surface)",
"can execute shell commands or code",
"untrusted input can reach code execution"
],
"tags": [
"code-exec",
"untrusted-input",
"sensitive-source",
"external-sink"
]
},
"coverage": {
"level": "source",
"inputs": {
"toolSurface": true,
"implementationSource": true,
"packageMetadata": true,
"liveTransport": false
},
"caveats": [
"Tools were statically extracted from the published source (35 recovered), not enumerated from a running server. Tool-poisoning, Unicode-smuggling, capability and toxic-flow analysis ran on this inferred surface, but a mis-parsed registration could be missed or mis-attributed, so tool-derived findings are capped below “confirmed”. To grade the real runtime surface, scan the running server: --command \"npx -y <package>\"."
]
},
"findings": [
{
"ruleId": "MTC-SRC-006",
"title": "Reads a sensitive credential path or dumps the environment (lib/profiles-presets.js)",
"category": "exfiltration",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`lib/profiles-presets.js:54`): 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.",
"remediation": "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": {
"kind": "server",
"name": "lib/profiles-presets.js"
},
"evidence": "fig', 'gcloud-home/.config/gcloud', ], }, aws: { // AWS isolation gotcha — config + credentials honour",
"owasp": "LLM02:2025 Sensitive Information Disclosure",
"data": {
"rule": "MTC-SRC-006",
"file": "lib/profiles-presets.js",
"line": 54,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-006",
"title": "Reads a sensitive credential path or dumps the environment (server.js)",
"category": "exfiltration",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`server.js:608`): 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.",
"remediation": "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": {
"kind": "server",
"name": "server.js"
},
"evidence": "\" \"\\$ANVIL_AWS_FAKE_HOME/.aws/credentials\" 2>/dev/null aws() { HOME=\"\\$ANVIL_AWS_FAKE_HOME\" __anvil_aws_inner \"\\$@\"; }",
"owasp": "LLM02:2025 Sensitive Information Disclosure",
"data": {
"rule": "MTC-SRC-006",
"file": "server.js",
"line": 608,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-006",
"title": "Reads a sensitive credential path or dumps the environment (src/client.js)",
"category": "exfiltration",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`src/client.js:1903`): 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.",
"remediation": "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": {
"kind": "server",
"name": "src/client.js"
},
"evidence": "ey: 'gcloud', label: '~/.config/gcloud' }, { key: 'aws', label: '~/.aws' }, { key: 'kubectl', label: '~/.kube'",
"owasp": "LLM02:2025 Sensitive Information Disclosure",
"data": {
"rule": "MTC-SRC-006",
"file": "src/client.js",
"line": 1903,
"nonRuntime": false
}
},
{
"ruleId": "MTC-CAP-001",
"title": "Tool \"terminal_run\" exposes command/code execution",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "Tool \"terminal_run\" appears to run shell commands or evaluate code (parameter \"command\"). Arbitrary execution driven by model input is one of the most dangerous MCP capabilities; combined with any untrusted input it becomes RCE.",
"remediation": "Sandbox execution, allowlist commands/arguments, and never pass model output to a shell unescaped.",
"location": {
"kind": "tool",
"name": "terminal_run"
},
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"tags": [
"code-exec"
]
}
},
{
"ruleId": "MTC-CAP-001",
"title": "Tool \"swarm_spawn\" exposes command/code execution",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "Tool \"swarm_spawn\" appears to run shell commands or evaluate code (keyword \"spawn\" in tool name). Arbitrary execution driven by model input is one of the most dangerous MCP capabilities; combined with any untrusted input it becomes RCE.",
"remediation": "Sandbox execution, allowlist commands/arguments, and never pass model output to a shell unescaped.",
"location": {
"kind": "tool",
"name": "swarm_spawn"
},
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"tags": [
"code-exec"
]
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (bin/anvil-showcase.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`bin/anvil-showcase.js:34151`): 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.",
"remediation": "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": {
"kind": "server",
"name": "bin/anvil-showcase.js"
},
"evidence": "lit(\"\\n\").map((line) => exec(expandTabs(line), columns, options)).join(\"\\n\"); } // node_modules/terminal-size/index.js",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "bin/anvil-showcase.js",
"line": 34151,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (bin/anvilterm-agent-loop.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`bin/anvilterm-agent-loop.js:24`): 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.",
"remediation": "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": {
"kind": "server",
"name": "bin/anvilterm-agent-loop.js"
},
"evidence": "st { spawn } = require('child_process'); const { setTimeout: delay } = require('timers/promises'); // --- Arg parsing -",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "bin/anvilterm-agent-loop.js",
"line": 24,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (bin/anvilterm-doctor.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`bin/anvilterm-doctor.js:24`): 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.",
"remediation": "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": {
"kind": "server",
"name": "bin/anvilterm-doctor.js"
},
"evidence": "spawnSync } = require('child_process'); const ROOT = path.resolve(__dirname, '..'); const SKILL_SRC = path.join(ROOT,",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "bin/anvilterm-doctor.js",
"line": 24,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (bin/anvilterm-mcp.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`bin/anvilterm-mcp.js:114`): 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.",
"remediation": "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": {
"kind": "server",
"name": "bin/anvilterm-mcp.js"
},
"evidence": "st { spawn } = require('child_process'); let spawnedServerProc = null; function readSwarmMemoryForPrompt() { try {",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "bin/anvilterm-mcp.js",
"line": 114,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (bin/anvilterm.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`bin/anvilterm.js:206`): 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.",
"remediation": "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": {
"kind": "server",
"name": "bin/anvilterm.js"
},
"evidence": "ync, spawn } = require('child_process'); // First positional non-flag arg is the path. Empty / no arg → no-op. cons",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "bin/anvilterm.js",
"line": 206,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (lib/arena-summary.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`lib/arena-summary.js:20`): 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.",
"remediation": "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": {
"kind": "server",
"name": "lib/arena-summary.js"
},
"evidence": "st { spawn } = require('child_process'); const ARENA_DIR = path.join(os.homedir(), '.anvil', 'arena'); const WORKSPACES",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "lib/arena-summary.js",
"line": 20,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (lib/codex-client.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`lib/codex-client.js:10`): 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.",
"remediation": "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": {
"kind": "server",
"name": "lib/codex-client.js"
},
"evidence": ", execFile } = require('child_process'); const { promisify } = require('util'); const pExecFile = promisify(execFile); c",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "lib/codex-client.js",
"line": 10,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (lib/forge.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`lib/forge.js:597`): 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.",
"remediation": "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": {
"kind": "server",
"name": "lib/forge.js"
},
"evidence": "ecFileSync } = require('child_process'); function fetchSkillRepo(item, destDir) { if (!item.repo) throw new Error('ski",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "lib/forge.js",
"line": 597,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (lib/keyring-store.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`lib/keyring-store.js:27`): 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.",
"remediation": "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": {
"kind": "server",
"name": "lib/keyring-store.js"
},
"evidence": "ecFileSync } = require('child_process'); const os = require('os'); const SERVICE_PREFIX = 'anvilterm'; function fullNam",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "lib/keyring-store.js",
"line": 27,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (lib/models-catalog.js)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`lib/models-catalog.js:10`): 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.",
"remediation": "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": {
"kind": "server",
"name": "lib/models-catalog.js"
},
"evidence": "{ execFile } = require('child_process'); const { promisify } = require('util'); const pExecFile = promisify(execFile);",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "lib/models-catalog.js",
"line": 10,
"nonRuntime": false
}
},
{
"ruleId": "MTC-FLOW-005",
"title": "Untrusted input can drive an external action",
"category": "exfiltration",
"severity": "medium",
"confidence": "strong",
"description": "Untrusted-input tools ([swarm_room_listen, swarm_room_thread]) co-exist with external-action tools ([terminal_run, swarm_spawn]). A prompt injection could cause unwanted external actions, though no direct sensitive-data leak path was found.",
"remediation": "Require confirmation for state-changing/egress actions triggered after processing untrusted content.",
"location": {
"kind": "flow",
"name": "swarm_room_listen → terminal_run"
},
"evidence": "untrusted [swarm_room_listen, swarm_room_thread] → sinks [terminal_run, swarm_spawn]",
"owasp": "LLM06:2025 Excessive Agency",
"references": [],
"data": {
"untrusted": [
"swarm_room_listen",
"swarm_room_thread"
],
"sinks": [
"terminal_run",
"swarm_spawn"
]
}
},
{
"ruleId": "MTC-SRC-003",
"title": "Hardcoded egress to an external endpoint (lib/models-catalog.js)",
"category": "exfiltration",
"severity": "medium",
"confidence": "strong",
"description": "In the server's implementation (`lib/models-catalog.js:75`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
"remediation": "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": {
"kind": "server",
"name": "lib/models-catalog.js"
},
"evidence": "0); const r = await fetch('https://models.dev/api.json', { signal: ac.signal }); clearTimeout(t); if (!r.ok)",
"owasp": "LLM02:2025 Sensitive Information Disclosure",
"data": {
"rule": "MTC-SRC-003",
"file": "lib/models-catalog.js",
"line": 75,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-003",
"title": "Hardcoded egress to an external endpoint (lib/usage-snapshot.js)",
"category": "exfiltration",
"severity": "medium",
"confidence": "strong",
"description": "In the server's implementation (`lib/usage-snapshot.js:58`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
"remediation": "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": {
"kind": "server",
"name": "lib/usage-snapshot.js"
},
"evidence": "y { const r = await fetch('https://chatgpt.com/backend-api/wham/usage', { headers: { Authorization: `B",
"owasp": "LLM02:2025 Sensitive Information Disclosure",
"data": {
"rule": "MTC-SRC-003",
"file": "lib/usage-snapshot.js",
"line": 58,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-003",
"title": "Hardcoded egress to an external endpoint (server.js)",
"category": "exfiltration",
"severity": "medium",
"confidence": "strong",
"description": "In the server's implementation (`server.js:4050`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
"remediation": "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": {
"kind": "server",
"name": "server.js"
},
"evidence": "ll; const r = await fetch('https://api.anthropic.com/api/oauth/usage', { headers: { 'Authorization': `",
"owasp": "LLM02:2025 Sensitive Information Disclosure",
"data": {
"rule": "MTC-SRC-003",
"file": "server.js",
"line": 4050,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-009",
"title": "Untrusted input concatenated into a command sink (bin/anvilterm-doctor.js)",
"category": "injection",
"severity": "medium",
"confidence": "strong",
"description": "In the server's implementation (`bin/anvilterm-doctor.js:55`): 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.",
"remediation": "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": {
"kind": "server",
"name": "bin/anvilterm-doctor.js"
},
"evidence": "n onPath(bin) { try { execSync(`command -v ${bin}`, { stdio: 'pipe' }); return true; } catch { return false; } } funct",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-009",
"file": "bin/anvilterm-doctor.js",
"line": 55,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-009",
"title": "Untrusted input concatenated into a command sink (bin/anvilterm.js)",
"category": "injection",
"severity": "medium",
"confidence": "strong",
"description": "In the server's implementation (`bin/anvilterm.js:323`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
"remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
"location": {
"kind": "server",
"name": "bin/anvilterm.js"
},
"evidence": "'); try { execSync(`npx esbuild \"${srcPath}\" --bundle --outfile=\"${bundlePath}\" --format=iife --platform=b",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-009",
"file": "bin/anvilterm.js",
"line": 323,
"nonRuntime": false
}
},
{
"ruleId": "MTC-CAP-006",
"title": "Unconstrained command parameter \"command\" on \"terminal_run\"",
"category": "permissions",
"severity": "medium",
"confidence": "heuristic",
"description": "Tool \"terminal_run\" takes a command-shaped parameter \"command\" with no enum/pattern constraint. Free-form, model- or attacker-controlled arguments reaching a shell is the command-injection precondition.",
"remediation": "Constrain the parameter (enum/pattern), or build the command from a fixed template with escaped args.",
"location": {
"kind": "tool",
"name": "terminal_run",
"field": "inputSchema.properties.command"
},
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"param": "command"
}
},
{
"ruleId": "MTC-SRC-005",
"title": "Dynamic module load from a non-literal (lib/codex-client.js)",
"category": "permissions",
"severity": "medium",
"confidence": "heuristic",
"description": "In the server's implementation (`lib/codex-client.js:190`): 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.",
"remediation": "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": {
"kind": "server",
"name": "lib/codex-client.js"
},
"evidence": "{ const mod = await import(SDK_PACKAGE); _Codex = mod.Codex; if (!_Codex) throw new Error('@openai/codex-sdk",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-005",
"file": "lib/codex-client.js",
"line": 190,
"nonRuntime": false
}
},
{
"ruleId": "MTC-SRC-006",
"title": "Reads a sensitive credential path or dumps the environment in packaging/dev tooling (public/bundle.js)",
"category": "exfiltration",
"severity": "low",
"confidence": "heuristic",
"description": "In a packaging/dev/install script (shipped, but not the server runtime) (`public/bundle.js:16979`): 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.",
"remediation": "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": {
"kind": "server",
"name": "public/bundle.js"
},
"evidence": "key: \"gcloud\", label: \"~/.config/gcloud\" }, { key: \"aws\", label: \"~/.aws\" }, { key: \"kubectl\", label: \"~/.kube\"",
"owasp": "LLM02:2025 Sensitive Information Disclosure",
"data": {
"rule": "MTC-SRC-006",
"file": "public/bundle.js",
"line": 16979,
"nonRuntime": true
}
},
{
"ruleId": "MTC-CAP-005",
"title": "Mutating tool \"terminal_run\" declares no destructiveHint",
"category": "hygiene",
"severity": "low",
"confidence": "heuristic",
"description": "Tool \"terminal_run\" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.",
"remediation": "Declare accurate annotations, and gate destructive tools on user confirmation regardless.",
"location": {
"kind": "tool",
"name": "terminal_run"
},
"data": {
"tags": [
"code-exec"
]
}
},
{
"ruleId": "MTC-CAP-005",
"title": "Mutating tool \"swarm_spawn\" declares no destructiveHint",
"category": "hygiene",
"severity": "low",
"confidence": "heuristic",
"description": "Tool \"swarm_spawn\" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.",
"remediation": "Declare accurate annotations, and gate destructive tools on user confirmation regardless.",
"location": {
"kind": "tool",
"name": "swarm_spawn"
},
"data": {
"tags": [
"code-exec"
]
}
},
{
"ruleId": "MTC-SRC-001",
"title": "Dynamic code execution in packaging/dev tooling (public/lottie.min.js)",
"category": "permissions",
"severity": "low",
"confidence": "heuristic",
"description": "In a packaging/dev/install script (shipped, but not the server runtime) (`public/lottie.min.js:1`): 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.",
"remediation": "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": {
"kind": "server",
"name": "public/lottie.min.js"
},
"evidence": "_rt,expression_function=eval(\"[function _expression_function(){\"+val+\";scoped_bm_rt=$bm_rt}]\")[0],numKeys=property.kf?da",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-001",
"file": "public/lottie.min.js",
"line": 1,
"nonRuntime": true
}
},
{
"ruleId": "MTC-SUP-010",
"title": "Package runs install-time scripts (postinstall)",
"category": "supply-chain",
"severity": "low",
"confidence": "heuristic",
"description": "\"anvilterm\" executes postinstall script(s) at install time. An install hook runs at install time; most are routine build/setup, but review what it does before trusting it.",
"remediation": "Review the scripts; install with --ignore-scripts where possible and vet what they do.",
"location": {
"kind": "package",
"name": "anvilterm"
},
"evidence": "node lib/postinstall.js",
"owasp": "LLM03:2025 Supply Chain",
"references": [
"https://github.com/ossf/package-analysis",
"https://owasp.org/www-project-top-10-ci-cd-security-risks/"
],
"data": {
"scripts": [
"postinstall"
]
}
},
{
"ruleId": "MTC-SUP-011",
"title": "Package declares no source repository",
"category": "supply-chain",
"severity": "info",
"confidence": "strong",
"description": "\"anvilterm\" declares no repository URL, so its published artifact cannot be compared against reviewable source.",
"remediation": "Prefer packages that link to public, reviewable source.",
"location": {
"kind": "package",
"name": "anvilterm"
}
}
],
"toxicFlows": [
{
"id": "flow-1",
"severity": "medium",
"confidence": "strong",
"untrustedInput": [
"swarm_room_listen",
"swarm_room_thread"
],
"sensitiveSource": [],
"externalSink": [
"terminal_run",
"swarm_spawn"
],
"selfContained": false,
"description": "Untrusted input can drive an external action even though no sensitive source is exposed."
}
],
"capabilities": [
{
"tool": "terminal_create",
"tags": [],
"reasons": {}
},
{
"tool": "terminal_write",
"tags": [],
"reasons": {}
},
{
"tool": "terminal_run",
"tags": [
"code-exec"
],
"reasons": {
"code-exec": [
"parameter \"command\""
]
}
},
{
"tool": "terminal_read",
"tags": [],
"reasons": {}
},
{
"tool": "terminal_screen",
"tags": [],
"reasons": {}
},
{
"tool": "terminal_screenshot",
"tags": [],
"reasons": {}
},
{
"tool": "terminal_list",
"tags": [],
"reasons": {}
},
{
"tool": "terminal_close",
"tags": [],
"reasons": {}
},
{
"tool": "tui_type",
"tags": [],
"reasons": {}
},
{
"tool": "tui_interrupt",
"tags": [],
"reasons": {}
},
{
"tool": "tui_choose",
"tags": [],
"reasons": {}
},
{
"tool": "tui_paste_ref",
"tags": [],
"reasons": {}
},
{
"tool": "swarm_spawn",
"tags": [
"code-exec"
],
"reasons": {
"code-exec": [
"keyword \"spawn\" in tool name"
]
}
},
{
"tool": "swarm_route",
"tags": [],
"reasons": {}
},
{
"tool": "swarm_vendors",
"tags": [],
"reasons": {}
},
{
"tool": "swarm_room_post",
"tags": [],
"reasons": {}
},
{
"tool": "swarm_room_listen",
"tags": [
"untrusted-input"
],
"reasons": {
"untrusted-input": [
"keyword \"fetch\""
]
}
},
{
"tool": "swarm_room_thread",
"tags": [
"untrusted-input"
],
"reasons": {
"untrusted-input": [
"keyword \"fetch\""
]
}
},
{
"tool": "swarm_room_list",
"tags": [],
"reasons": {}
},
{
"tool": "swarm_memory_read",
"tags": [],
"reasons": {}
},
{
"tool": "swarm_memory_update",
"tags": [],
"reasons": {}
},
{
"tool": "arena_push_artifact",
"tags": [],
"reasons": {}
},
{
"tool": "arena_current",
"tags": [],
"reasons": {}
},
{
"tool": "swarm_artifact_save",
"tags": [],
"reasons": {}
},
{
"tool": "analyze_video_or_youtube",
"tags": [],
"reasons": {}
},
{
"tool": "video_analyze",
"tags": [],
"reasons": {}
},
{
"tool": "lookup_models",
"tags": [],
"reasons": {}
},
{
"tool": "arena_estimate_cost",
"tags": [],
"reasons": {}
},
{
"tool": "arena_launch",
"tags": [],
"reasons": {}
},
{
"tool": "image_battle_launch",
"tags": [],
"reasons": {}
},
{
"tool": "swarm_check_stuck",
"tags": [],
"reasons": {}
},
{
"tool": "task_list",
"tags": [],
"reasons": {}
},
{
"tool": "task_create",
"tags": [],
"reasons": {}
},
{
"tool": "task_update",
"tags": [],
"reasons": {}
},
{
"tool": "task_delete",
"tags": [],
"reasons": {}
}
],
"surfaceDigest": "f126001bce5983b0386e1a1311b44133e5cecc8aae8d374455c268f1d7021be9",
"stats": {
"tools": 35,
"prompts": 0,
"resources": 0,
"findingsBySeverity": {
"critical": 0,
"high": 15,
"medium": 8,
"low": 5,
"info": 1
}
}
}