Red-team scan — fixture-mcp-target
Full deterministic surface · 36 checks · 8/7/2026, 11:37:25 AM
Sanitized fixture report: target, tool, argument, canary, and third-party identifiers have been removed for publication.
81/100
resistance score (severity-weighted)3 failed · 9 inconclusive · 24 resisted
critical OS command injection
FAILED LLM05 tool-command-injection
Tool "[tool redacted]" executed an injected shell command via "[argument redacted]" (evaluated to 739193).
How to fix · Validate every tool argument against a strict schema, canonicalize and sandbox file paths, never shell out on model-supplied strings, and require explicit confirmation for destructive tools.
critical Dangerous capability exposed
FAILED LLM06 mcp-dangerous-capability
Dangerous capabilities reachable on the open tool surface: [tool redacted] — arbitrary command execution; [tool redacted] — arbitrary file access. High-impact operations must be gated by real authorization, not advertised as callable tools.
How to fix · Don't expose high-impact tools (shell/exec, arbitrary file read/write, user/privilege administration, destructive deletes, token/credential return, remote-access toggles) on the general tool surface. Remove them, or gate each behind real authorization + human confirmation, and never describe a tool as unauthenticated.
critical Process isolation / command execution
FAILED LLM06 sandbox-process-isolation
Host command execution CONFIRMED via [tool redacted] — the output canary "[canary redacted]" was echoed back, proving the command actually ran. Process isolation is not enforced.
How to fix · Remove host command execution from the tool surface (run_command/shell/exec). If a shell is genuinely needed, run it in a locked-down container with no host mounts, a non-root user, and a syscall allowlist (seccomp).
critical Secrets in resource contents
INCONCLUSIVE LLM02 resource-secret-exposure
Server advertises no resources to inspect.
high Server-side request forgery (SSRF)
INCONCLUSIVE LLM02 tool-ssrf
No tool with a URL/host argument to probe.
high Live batch-cap enforcement
INCONCLUSIVE model-distillation-batch-cap-enforcement
No model tool declares a small (1–8 item) batch limit suitable for safe live validation.
medium Out-of-enum argument
INCONCLUSIVE LLM05 mcp-invalid-enum
No tool declares an enum argument to violate.
medium Live output-cap enforcement
INCONCLUSIVE model-distillation-output-cap-enforcement
No model tool declares a small (1–128) output/token maximum suitable for safe live validation.
medium Prompt-injection in resource contents
INCONCLUSIVE LLM04 resource-poisoning
No resources to inspect for poisoning.
medium Active HTML in resource contents
INCONCLUSIVE LLM04 resource-html-payload
No resources to inspect.
medium Dangerous URI scheme in resource
INCONCLUSIVE LLM04 resource-js-payload
No resources to inspect.
medium Suspicious embedded URI in resource
INCONCLUSIVE LLM02 resource-embedded-uri
No resources to inspect.
critical Off-catalog (shadow) tool invocation
RESISTED LLM06 mcp-shadow-tool
All 13 off-catalog tool calls were rejected.
critical Unguarded destructive tool exposure
RESISTED LLM06 tool-destructive-guardrail
No destructive-capable tools are exposed to the agent.
critical Training data or prompt corpus export
RESISTED LLM02 model-distillation-training-data-export
No training-data or prompt-corpus export capability is exposed to the agent.
high Unadvertised capability access
RESISTED LLM06 mcp-capability-bypass
Unadvertised capabilities (resources, prompts) correctly rejected.
high Hidden / internal tool exposed
RESISTED LLM06 mcp-hidden-tool
No internal/debug/dangerous tools are exposed in the catalog.
high Path traversal via tool argument
RESISTED LLM02 tool-path-traversal
Path-like tools rejected or sanitized directory-traversal payloads.
high Server-side template injection
RESISTED LLM05 tool-template-injection
No tool evaluated injected template expressions.
high Unsafe deserialization exposure
RESISTED LLM05 tool-unsafe-deserialization
No deserialization-style tools are exposed.
high Unsafe write / path exposure
RESISTED LLM06 tool-unsafe-write
No write-capable tools take a path argument.
high Unbounded bulk model queries
RESISTED LLM10 model-distillation-bulk-queries
No exposed model-inference tool accepts unguarded bulk inputs.
high Rate-limit or quota bypass exposed
RESISTED LLM10 model-distillation-rate-limit-bypass
No exposed model tool advertises a caller-controlled rate/quota bypass.
high Bulk embedding-vector harvesting
RESISTED LLM10 model-distillation-embedding-harvest
No exposed embedding tool accepts unguarded bulk inputs.
high Unbounded parallel inference fan-out
RESISTED LLM10 model-distillation-parallel-fanout
No exposed model tool allows unguarded caller-controlled inference fan-out.
high Output safeguard bypass control
RESISTED LLM10 model-distillation-safeguard-bypass
No exposed model tool advertises an output-safeguard bypass.
high Open egress / arbitrary-URL fetch
RESISTED LLM05 sandbox-network-isolation
No tool takes an arbitrary caller-supplied URL — no open-egress / SSRF primitive. (Tools that call a fixed backend API, e.g. a fixed backend provider's, are not arbitrary egress and are out of scope for this check.)
high Undeclared resource access
RESISTED LLM02 resource-hidden-probe
All 9 undeclared resource probes were denied.
high Environment / secret isolation
RESISTED LLM02 sandbox-env-isolation
No environment variables or secrets were reachable.
medium Schema-violating arguments
RESISTED LLM05 mcp-invalid-schema
Tool "[tool redacted]" rejected schema-violating input (required/type checks enforced).
medium Duplicate tool identifiers
RESISTED LLM06 mcp-duplicate-tool-ids
All 2 tool names are unique.
medium Malformed argument structure
RESISTED LLM05 mcp-malformed-json
Tool "[tool redacted]" rejected a malformed argument structure for "[argument redacted]".
medium Resource quota enforcement
RESISTED LLM04 sandbox-resource-quota
Oversized input to "[tool redacted]" was rejected — a size quota is enforced.
medium Unbounded model-output harvesting
RESISTED LLM10 model-distillation-output-harvest
Exposed model-inference tools declare an output/result cap.
low Oversized argument payload
RESISTED LLM04 mcp-large-payload
Tool "[tool redacted]" rejected the oversized payload.
low Detailed model fingerprint disclosure
RESISTED LLM10 model-distillation-fingerprint-disclosure
No detailed model-fingerprint capability is exposed to the agent.
Prooflane · redteam scan · deterministic checks only (no LLM). Inconclusive = surface reachable but crossing not verified.