mcp-tool-name-integrity-v1
5 rules · Keryn Software Licence · rules 0.8.0 · all packs
How to read a rule
- Severity
- How bad the finding is if it is real:
errorfails a default gate,warningandinfodo not. - Confidence
- P0 exploitable as written; P1 dangerous in context; P2 a heuristic or a posture check: confirm by reading the code.
- False-positive risk
- low: an exact API or flag, so hits are almost always real; medium: the pattern is legitimate in many programs, so check each hit; high: it finds a lead to confirm, so expect benign hits.
- Edge
- The relationship the finding sits on. reaches: content or traffic gets to a thing; calls: one thing invokes another; reads: one thing reads from another; writes: one thing writes to another; sends: data leaves for another party; grants: authority is given or left unchecked; depends_on: something relies on a component it does not verify; trusts: one party believes another without evidence. The boundary it crosses, when there is one: user->prompt (untrusted input reaching a prompt); model->code (model output reaching code, a shell, a query or a page); secret->outside (a secret leaving the trusted side); supply-chain (code, models or prompts pulled in from outside).
- Taxonomy chips
- References for readers who work in OWASP LLM 2025, OWASP Agentic, CWE or NIST AI RMF. They say where a rule belongs, not that a framework is satisfied.
seceng.mcp.py.tool-name-non-ascii
Section titled “seceng.mcp.py.tool-name-non-ascii”error python mcp, python, fastmcp
This MCP tool is registered under a name that contains characters outside printable ASCII (a homoglyph, a zero-width or other invisible code point). Tool names are matched as identifiers by clients and by the model; a name that looks like a trusted tool but is a different string lets this tool capture calls meant for the other one, or hide from allow-lists that compare names. MCP tool names are plain ASCII identifiers.
Confidence P1: dangerous in context. False-positive risk low: an exact API or flag, so hits are almost always real.
Edge inspected: calls one thing invokes another.
CWE-1007 CWE-451 LLM06 ASI04 ASI03
Fix: Name tools with ASCII letters, digits, _ and - only, reject any other code point at registration, and have the client refuse or namespace tools whose name collides with or confusably resembles one already registered.
References: https://modelcontextprotocol.io/specification/2025-06-18/server/tools · https://modelcontextprotocol.io/docs/2025-11-25/tutorials/security/security_best_practices · https://www.unicode.org/reports/tr39/
seceng.mcp.ts.tool-name-non-ascii
Section titled “seceng.mcp.ts.tool-name-non-ascii”error typescript, javascript mcp, typescript, nodejs
This MCP tool is registered under a name that contains characters outside printable ASCII (a homoglyph, a zero-width or other invisible code point). Tool names are matched as identifiers by clients and by the model; a name that looks like a trusted tool but is a different string lets this tool capture calls meant for the other one, or hide from allow-lists that compare names. MCP tool names are plain ASCII identifiers.
Confidence P1: dangerous in context. False-positive risk low: an exact API or flag, so hits are almost always real.
Edge inspected: calls one thing invokes another.
CWE-1007 CWE-451 LLM06 ASI04 ASI03
Fix: Name tools with ASCII letters, digits, _ and - only, reject any other code point at registration, and have the client refuse or namespace tools whose name collides with or confusably resembles one already registered.
References: https://modelcontextprotocol.io/specification/2025-06-18/server/tools · https://modelcontextprotocol.io/docs/2025-11-25/tutorials/security/security_best_practices · https://www.unicode.org/reports/tr39/
seceng.mcp.py.tool-name-lookalike-of-common-tool
Section titled “seceng.mcp.py.tool-name-lookalike-of-common-tool”warning python mcp, python, fastmcp
This MCP tool name is a case or separator variant of a widely deployed tool name (the reference filesystem, fetch, GitHub and shell servers) without being that name. Clients and models that pick tools by name confuse readFile, Read_File or read-file with read_file; a server that ships such a near-miss either collides with the well-known tool by accident or is positioned to capture its calls. Use the canonical spelling if this is that tool, or a clearly distinct, namespaced name if it is not.
Confidence P2: a heuristic or a posture check: confirm by reading the code. False-positive risk medium: the pattern is legitimate in many programs, so check each hit.
Edge inspected: calls one thing invokes another.
Fix: Either adopt the canonical name of the tool you are reimplementing, or prefix the name with your server’s namespace (acme_read_file) so no client or model can mistake it for the reference tool.
References: https://modelcontextprotocol.io/specification/2025-06-18/server/tools · https://github.com/modelcontextprotocol/servers
seceng.mcp.ts.tool-name-lookalike-of-common-tool
Section titled “seceng.mcp.ts.tool-name-lookalike-of-common-tool”warning typescript, javascript mcp, typescript, nodejs
This MCP tool name is a case or separator variant of a widely deployed tool name (the reference filesystem, fetch, GitHub and shell servers) without being that name. Clients and models that pick tools by name confuse readFile, Read_File or read-file with read_file; a server that ships such a near-miss either collides with the well-known tool by accident or is positioned to capture its calls. Use the canonical spelling if this is that tool, or a clearly distinct, namespaced name if it is not.
Confidence P2: a heuristic or a posture check: confirm by reading the code. False-positive risk medium: the pattern is legitimate in many programs, so check each hit.
Edge inspected: calls one thing invokes another.
Fix: Either adopt the canonical name of the tool you are reimplementing, or prefix the name with your server’s namespace (acme_read_file) so no client or model can mistake it for the reference tool.
References: https://modelcontextprotocol.io/specification/2025-06-18/server/tools · https://github.com/modelcontextprotocol/servers
seceng.mcp.config.server-package-lookalike-official-scope
Section titled “seceng.mcp.config.server-package-lookalike-official-scope”error generic mcp, npm, agent-config
This MCP client configuration installs a server package from an npm scope that imitates @modelcontextprotocol (the scope of the reference servers) but is not it. A lookalike scope is how a typosquatted server gets launched with the user’s credentials and filesystem; the package runs on every client start. Install reference servers from @modelcontextprotocol/... only, pinned to an exact version, and review the publisher of anything else.
Confidence P1: dangerous in context. False-positive risk low: an exact API or flag, so hits are almost always real.
Edge inspected: depends_on something relies on a component it does not verify; it crosses the supply-chain boundary (code, models or prompts pulled in from outside).
Fix: Replace the scope with @modelcontextprotocol if the reference server was meant, pin an exact version (@scope/pkg@1.2.3), and check the publisher and download counts of any other scope before launching it from an agent.
References: https://github.com/modelcontextprotocol/servers · https://docs.npmjs.com/about-scopes · https://modelcontextprotocol.io/docs/2025-11-25/tutorials/security/security_best_practices