ClawHub
Back to skill

Security audit

RMBase

Security checks for vulnerabilities and agentic risk

Overview

This RMBase skill is a coherent scientific lookup tool with disclosed public-network access, local caching, and explicit safeguards around uploads.

Install only if you are comfortable with RMBase queries and authorized public data being sent over unencrypted HTTP to bioinformaticsscience.cn. Use offline mode for private local work, avoid uploading unpublished BED or gene-list data, and review cached data in ~/.cache/rmbase if local persistence matters.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (5)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description frames the capability as querying RMBase with bundled scripts and official datasets, but the documented behavior also includes live HTTP access to the public site, POST submissions to upload endpoints, and generic transport/caching infrastructure. This mismatch can cause operators or policy engines to trust the skill more than warranted, leading to unanticipated transmission of user data over public HTTP or use of broader network functionality than expected.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding

The skill declares no explicit tool/permission scope even though its documented behavior includes reading and writing cache/files and making network requests. Without an allowlist or permissions declaration, an agent platform may grant broader capabilities than intended, increasing the chance of unintended data access, persistence, or exfiltration through the bundled Python workflow.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill enables implicit invocation without any trigger phrases or scope constraints, which can cause the agent to call the RMBase tool when a user did not explicitly request it. That increases the chance of unintended data access, prompt-surface expansion, or tool use based on ambiguous prompts, especially in multi-tool environments where provenance and query parameters matter.

Content

No source excerpt is available for this finding.

Unbounded Resource Access

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.

Content

Scanner excerpt · scripts/client.py (reported line 36)May include surrounding context.

python
def identifier(value, label="identifier"):
    if not isinstance(value, str) or not re.fullmatch(r"[A-Za-z0-9_][A-Za-z0-9_.:+-]{0,159}", value) or value.lower() == "all":
        raise RMBaseError("invalid_identifier", f"Provide a single explicit {label}; wildcards and unbounded queries are not supported.")
    return value

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · scripts/rmbase.py (reported line 119)May include surrounding context.

python
if cmd == "sites":
        return client.sites(args.type, args.cell, args.min_support, args.max_support)
    if cmd in {"enzyme", "cluster", "snorna"}:
        value = getattr(args, {"enzyme": "enzyme", "cluster": "type", "snorna": "name"}[cmd])
        return client.lookup(cmd, value)
    if cmd == "motif":
        if args.datasets and args.type:

Static analysis

No suspicious patterns detected.