Credenzo GitHub

mcp-audit is becoming Credenzo, the taste test for agent tools. The package is still mcp-audit-scanner and the command is still mcp-audit.

What mcp-audit checks

mcp-audit can report 225 finding IDs, plus 6 families of IDs generated at scan time. They come from 8 built-in analyzers, 40 community rules that run by default and 91 Semgrep rules for MCP server source code.

Severities come from the severity framework and the rule files; OWASP codes from the mapping file.

Repository files that run code 8 checks

Files a cloned repository can use to start programs when you open or trust it: project MCP configs, VS Code tasks that run when the folder opens, auto-run files for Gemini CLI, Codex and Claude Code, git settings and hooks that run a program when an agent runs git, and symlinks that point somewhere else.

Runs with mcp-audit scan --project .

IDWhat it flagsSeverityOWASP MCP
TRUST-001MCP server defined in project-level config (TrustFall risk)HIGHMCP09
TRUST-002Project-scoped config path is a symlinkHIGH, MEDIUM or INFOMCP09 MCP05
TRUST-003Repo-planted IDE auto-execution fileCRITICAL or HIGHMCP05 MCP09
TRUST-004Config path is a symlink (user-global / explicit)INFOMCP09
TRUST-005Symlinked directory was not scannedLOWMCP09
TRUST-006Git config runs a commandHIGH, MEDIUM, LOW or INFOMCP05
TRUST-007Git repository shape runs repo-controlled codeHIGH or INFOMCP05
TRUST-008Repo-planted auto-run file of another coding-agent harnessCRITICAL, HIGH or MEDIUMMCP05 MCP09

AI agents in CI workflows 3 checks

GitHub Actions workflows where an AI coding agent step reads text a stranger can write, such as an issue or a comment, while it holds secrets, write access and tools that can run commands. Workflow files are parsed as data and never run.

Runs with mcp-audit check mcp-audit scan --project .

IDWhat it flagsSeverityOWASP MCP
WORKFLOW-001CI agent runs on text a stranger can write, with secrets, write access and unrestricted toolsHIGHMCP06 MCP01
WORKFLOW-002CI agent's only tool restriction is a blocklistMEDIUMMCP06 MCP01
WORKFLOW-003Gemini CLI step exposed to the CVSS 10.0 workspace-trust flaw (GHSA-wpqr-6v78-jr5g)HIGHMCP05 MCP04

Coding-agent settings and permission rules 39 checks

Claude Code settings that skip permission prompts, approve every project server, or pre-approve commands that can write or reach the network. Permission rules that Claude Code ignores or that are not a boundary. Sandbox settings a repository can loosen. Hook commands that reach the network or touch agent config files. Servers that your organisation's managed Claude Code policy would block, policy set only where the user can change it, safety settings that got weaker since the last scan, editor extension allowlists that are not enforced, and credential files a session can read without a prompt.

Runs with mcp-audit check mcp-audit scan --project .

IDWhat it flagsSeverityOWASP MCP
CFHYG-005Claude Code hooks section detected in configMEDIUMMCP01 MCP07
EXTPOL-001Editor extension allowlist inventoryINFOMCP04
EXTPOL-002Editor extension allowlist is set only in user settingsLOWMCP04
EXTPOL-003Editor extension allowlist policy allows everythingLOWMCP04
GUARD-001Agent guardrail loosened since the last scanHIGH or MEDIUMMCP02
GUARD-002Managed settings file changed since the last scanINFOMCP02
HARNESS-001Claude Code sessions start with permission prompts skippedHIGH or MEDIUMMCP02 MCP07
HARNESS-002Every project .mcp.json server is approved without a promptHIGHMCP09 MCP07
HARNESS-003Confirmation before bypass mode is switched offMEDIUMMCP02 MCP07
HARNESS-004Every shell command is pre-approvedHIGHMCP02 MCP05
HARNESS-005Allow rules pre-approve commands that can write or reach the networkMEDIUMMCP02 MCP05
HARNESS-006Approval is off and nothing guards recursive deletesMEDIUMMCP02 MCP05
HARNESS-007Named .mcp.json servers are approved without a promptINFOMCP09 MCP07
HOOK-001Hook command contains network-egress instructionHIGHMCP05 MCP07
HOOK-002Hook command references agent config file pathHIGHMCP05 MCP07
HOOK-004Guard hook command may never run on WindowsMEDIUM or LOWMCP02 MCP07
HOOK-005Cursor hooks file can't be read, so its hooks never runMEDIUMMCP02 MCP07
MANAGED-001Managed MCP policy on this machineINFOMCP09
MANAGED-002Configured MCP server that the managed policy keeps from loadingINFOMCP09
MANAGED-003Managed settings are deployed but set no MCP policyLOWMCP09
MANAGED-004Settings file adds servers to an unlocked managed MCP allowlistMEDIUM or LOWMCP02 MCP09
MANAGED-005Managed MCP allowlist admits servers by name onlyLOWMCP09 MCP07
MANAGED-006Claude Code policy is set, not enforcedMEDIUMMCP09 MCP07
MANAGED-007Claude Code admin policy source is not a valid settings objectINFOMCP09
PERM-001Permission rules that Claude Code ignores or never matchesHIGH, MEDIUM or INFOMCP07 MCP02
PERM-002Path rules with one leading slash point inside the project, not at /MEDIUM or INFOMCP07 MCP10
PERM-003Allow rules whose trailing * also matches other programsLOWMCP02 MCP05
PERM-004Deny or ask rules on network or delete commands are not a boundary without the sandboxMEDIUM or INFOMCP05 MCP07
PERM-005WebFetch(domain:*) allow opens the sandbox to every hostHIGH or LOWMCP02 MCP10
PERM-006A bare Bash ask rule is skipped for sandboxed commandsMEDIUMMCP07 MCP05
PERM-007Allow rule pre-approves every tool of a server with high-severity findingsHIGHMCP02 MCP07
REACH-001Credential files a Claude Code session can read without askingLOWMCP01
REACH-002Claude Code's login store is readable by other local accountsMEDIUMMCP01
REACH-003.env files in the working directory with no Read deny ruleLOWMCP01
SANDBOX-001Repository settings weaken or switch off the Bash sandboxHIGH, MEDIUM or INFOMCP02 MCP05
SANDBOX-002Repository settings take code-running commands out of the sandboxHIGH or MEDIUMMCP05 MCP02
SANDBOX-003Repository settings let sandboxed commands reach hosts anyone can publish toMEDIUM or INFOMCP02 MCP10
SANDBOX-004Sandbox keys this settings file can't set (ignored by Claude Code)INFOMCP02
SANDBOX-005Local Claude Code settings are not ignored by gitMEDIUM or LOWMCP10 MCP02

Agent instruction files, skills and subagents 22 checks

Text the agent reads as instructions: Claude Code commands, CLAUDE.md and AGENTS.md files, the files they import, Cursor rules, and Copilot instruction and prompt files. mcp-audit looks for injected orders, hidden characters, outside links and heartbeat files that hide a step. It reads every file in a Claude Code skill, not only SKILL.md, for install steps that run remote code. It reads subagent definitions for looser permissions, hooks and their own MCP servers, and reports when an instruction file changed since the last scan.

Runs with mcp-audit scan --project . --include-agent-files mcp-audit agent-files scan

IDWhat it flagsSeverityOWASP MCP
AGENT-001Subagent asks for a looser permission modeMEDIUM or LOWMCP02 MCP07
AGENT-002disallowedTools specifier removes the whole toolINFOMCP02
AGENT-003Subagent defines hooks that run shell commandsMEDIUM or INFOMCP05 MCP07
AGENT-004Subagent declares its own MCP serverMEDIUM, LOW or INFOMCP09
INSTR-001Instruction file changed since the last scanINFOMCP10
INSTR-002Instruction file changed and the change brought a findingMEDIUMMCP10 MCP03
MEM-001Imperative override instruction in memory fileMEDIUMMCP03
MEM-002Injection pattern in memory fileMEDIUMMCP03
MEM-003Instruction file imports a file from outside the project / a .gitignore'd fileMEDIUM or LOWMCP10
MEM-004Heartbeat file hides what the agent doesHIGHMCP06 MCP08
MEM-005Instruction file holds a conditional sleeper triggerMEDIUMMCP03 MCP06
MEM-006Instruction file tells the agent to send secrets outHIGH or MEDIUMMCP03 MCP01
MEM-007Agent instruction file in Downloads or DesktopMEDIUMMCP03 MCP06
MEM-008Instruction file changes machine-wide agent settings or approvalsHIGH or MEDIUMMCP03 MCP02
SKILL-001Injection/exfiltration instruction in skill fileHIGH or INFOMCP03 MCP01
SKILL-002Obfuscated or oversized block in skill/command fileMEDIUMMCP03
SKILL-003Skill file references external URLMEDIUMMCP03 MCP04
SKILL-004Skill bundles executable scriptsINFOMCP03
SKILL-005Skill contains an install step that runs remote or hidden codeHIGH, LOW or INFOMCP04 MCP05
SKILL-006Skill ships something Claude Code runs, or tells the agent to drop its guardrailsHIGH, MEDIUM or LOWMCP01 MCP02 MCP04 MCP05 MCP07 MCP09
SKILL-007Skill contains content that cannot be inspectedMEDIUM or LOWMCP03 MCP04
SKILL-008Skill bundles code and has another findingHIGHMCP04 MCP05

Claude Code plugins and mods 10 checks

Installed plugins whose recorded version or checkout does not match their marketplace pin, or that follow an unpinned branch, and mods that can approve tool calls, read secrets and send data out, change what Claude reads, or start programs. Mods are read, never run.

Runs with mcp-audit check mcp-audit plugins scan

IDWhat it flagsSeverityOWASP MCP
MOD-001Mod can approve tool calls before the permission promptHIGHMCP07 MCP04
MOD-002Mod can read secrets or files and send data outHIGH or MEDIUMMCP01 MCP04
MOD-003Mod can change what Claude readsMEDIUMMCP03 MCP06
MOD-004Mod starts programs or changes their environmentMEDIUMMCP05 MCP04
MOD-005Claude Code mod installedINFOMCP04
MOD-006Plugin directories load in every Claude Code sessionMEDIUMMCP04
PLUGIN-001Installed plugin does not match its marketplace pinMEDIUMMCP04
PLUGIN-002Plugin source is not pinned to a commitLOWMCP04
PLUGIN-003Plugin checkout is not the commit its marketplace pinsCRITICAL or MEDIUMMCP04
PLUGIN-004Plugin update channel cannot hold its pinHIGH, MEDIUM or LOWMCP04

MCP server configs 20 checks

How each configured server is launched and reached: secrets written into the config, plain-HTTP endpoints, sudo launches, unpinned package fetches, remote servers with no authentication, and config files other users can read or write. If a check crashes on a server, the scan reports it instead of passing that server as clean.

Runs with mcp-audit check

IDWhat it flagsSeverityOWASP MCP
AUTH-001Remote MCP server configured without authenticationHIGH or MEDIUMMCP06
AUTH-002OAuth-configured server missing audience/resource bindingMEDIUMMCP06
CFHYG-001MCP config file is world-readableHIGHMCP01
CFHYG-002MCP config file resides in world-writable directoryHIGHMCP01 MCP09
CFHYG-003Config file stores inline plaintext secretHIGHMCP01
CFHYG-004Config uses environment variable references for credentialsINFOMCP01
CFHYG-006ANTHROPIC_BASE_URL overrides Anthropic API endpointMEDIUMMCP01 MCP08
CFHYG-007Project settings env redirects API trafficHIGHMCP01 MCP08
CFHYG-008Project settings send API requests to an Azure Foundry resourceINFOMCP01 MCP08
CRED-001API key or secret in environment variable valueHIGHMCP01
CRED-002API key or secret in command argumentsHIGHMCP01
CRED-003Literal secret in authentication headerHIGH or INFOMCP01
CRED-004Secret saved in a permission ruleHIGH or MEDIUMMCP01
CRED-005Secret in a permission rule in git historyHIGHMCP01
CRED-006Claude Code settings file copied into a container image or npm packageHIGH or LOWMCP01
SCAN-ERRAnalyzer crashed on a serverHIGHnone
TRANSPORT-001Unencrypted remote endpoint (HTTP)MEDIUMMCP07
TRANSPORT-002Elevated privilege execution (sudo/root)HIGHMCP02 MCP05
TRANSPORT-003Runtime package fetching without version pinMEDIUM or LOWMCP04
TRANSPORT-004Server binds to wildcard network interface (0.0.0.0)HIGHMCP07

Tool descriptions 29 checks

Orders hidden in the descriptions an agent reads: exfiltration instructions, prompt-format injection, behaviour overrides, invisible Unicode and lookalike characters. With --connect it reads live tool lists, flags a tool name claimed by two servers, and flags descriptions that name another server's tool or tell the agent to rewrite other tools' arguments. It also flags a tool marked read-only whose own inputs run commands or write.

Runs with mcp-audit check mcp-audit scan --connect

IDWhat it flagsSeverityOWASP MCP
ANNOT-001Tool annotation contradicts the tool's own input schemaLOWMCP03 MCP06
COLLIDE-001Tool name collision: tool claimed by multiple serversHIGH or MEDIUMMCP01 MCP09
DELEG-001Tool description names another server's toolMEDIUMMCP03 MCP06
DELEG-002Tool description tells the agent to act when something else happensMEDIUMMCP03 MCP06
DELEG-003Tool description tells the agent to rewrite other tools' argumentsHIGHMCP03 MCP06
POISON-001Tool description contains SSH/credentials exfiltration instructionsCRITICALMCP03 MCP01
POISON-002Tool description contains cloud credentials exfiltration instructionsCRITICALMCP03 MCP01
POISON-003Tool description contains .env file exfiltration instructionsCRITICALMCP03 MCP01
POISON-010Tool description contains XML instruction injectionHIGHMCP03 MCP06
POISON-011Tool description contains LLM prompt format injection markersHIGHMCP03 MCP06
POISON-012Tool description contains behavioral override instructionHIGHMCP03 MCP06
POISON-020Tool description contains data exfiltration via encoding or side channelHIGHMCP03 MCP10
POISON-021Tool description contains hidden parameter exfiltrationHIGHMCP03 MCP10
POISON-030Tool description contains cross-tool manipulation instructionMEDIUMMCP03 MCP06
POISON-040Tool description contains zero-width Unicode charactersMEDIUMMCP03
POISON-041Concealed instruction via concealment channelHIGHMCP03 MCP01
POISON-042Concealment channel present, no matched payloadLOWMCP03
POISON-043Zero-width binary channel present, no matched payloadMEDIUMMCP03
POISON-044Variation-selector channel present, no matched payloadMEDIUMMCP03
POISON-045Text hidden below blank-line paddingHIGH or MEDIUMMCP03
POISON-050Excessive tool description lengthLOWMCP03
POISON-060Tool description contains Unicode homoglyph charactersHIGHMCP03
POISON-070Secrecy directiveHIGHMCP03 MCP06
POISON-071Autonomy override directiveHIGH or LOWMCP03 MCP06
POISON-072Ask-the-user tool disabledMEDIUMMCP03 MCP06
POISON-073Coercive framingLOWMCP03 MCP06
POISON-074Startup execution directiveLOWMCP03 MCP05
POISON-075Remote instruction followingLOWMCP03 MCP06
POISON-076Oversight bypass directiveHIGHMCP03 MCP06

Dangerous combinations of servers 12 checks

Servers that are harmless alone but together let an injected prompt read data and send it out, including by posting to GitHub, Slack or Jira, or write a file another server can run. killchain ranks the changes that break the most paths.

Runs with mcp-audit check mcp-audit killchain

IDWhat it flagsSeverityOWASP MCP
INTEG-001Dangerous capability combination: file write + shell execution (plant-then-execute)HIGHMCP05
TOXIC-001Dangerous capability combination: file read + network exfiltrationHIGHMCP05 MCP10
TOXIC-002Dangerous capability combination: file read + email exfiltrationHIGHMCP05 MCP10
TOXIC-003Dangerous capability combination: secrets store + network accessCRITICALMCP01 MCP10
TOXIC-004Dangerous capability combination: file read + shell executionHIGHMCP10
TOXIC-005Dangerous capability combination: database + networkHIGHMCP05 MCP10
TOXIC-006Dangerous capability combination: shell execution + networkCRITICALMCP05 MCP10
TOXIC-007Capability combination: git access + networkMEDIUMMCP10
TOXIC-008Dangerous capability combination: file read + comms postingHIGHMCP05 MCP10
TOXIC-009Dangerous capability combination: secrets store + comms postingCRITICALMCP01 MCP10
TOXIC-010Dangerous capability combination: database + comms postingHIGHMCP05 MCP10
TOXIC-011Dangerous capability combination: shell execution + comms postingCRITICALMCP05 MCP10

Packages and supply chain 19 checks

Lookalike package names, unscoped twins of known servers, packages listed as malicious in OSV, recorded advisories for a package, deprecated versions, and missing or mismatched Sigstore provenance. Provenance, hash and CVE checks use the network, so they are opt-in.

Runs with mcp-audit vet <package> mcp-audit scan --verify-signatures

IDWhat it flagsSeverityOWASP MCP
ATTEST-010Sigstore provenance verified (valid match)INFOnone
ATTEST-011Sigstore attestation from unexpected publisherHIGHMCP04
ATTEST-012Invalid Sigstore signature (tampered or forged)CRITICALMCP04
ATTEST-013Expected Sigstore attestation missing for known-provenance packageMEDIUMMCP04
ATTEST-014No Sigstore attestation foundMEDIUM or INFOMCP04
ATTEST-015Sigstore attestation verification errorINFOnone
SC-001Package name is one edit distance from a known-good package (typosquat)CRITICALMCP04
SC-002Package name is two edits from a known-good packageHIGHMCP04
SC-003Package name is three edits from a known-good packageMEDIUMMCP04
SC-004Known CVE advisory recorded for this packageHIGH or MEDIUMMCP04
SC-005Deprecated package resolved at lock timeMEDIUMMCP04
SC-007Unscoped twin of a known packageMEDIUMMCP04
SC-008Package listed as malicious in OSV (MAL-* record)CRITICAL or MEDIUMMCP04
SC-009Known CVE fix not comparable with pinned versionINFOMCP04
SC-010Archived MCP server, no security fixes comingMEDIUMMCP04
SC-011Known malicious package in the npx cacheCRITICALMCP04
VULN-UNPINNEDUnpinned version (scan --check-vulns)LOWMCP04
VULN-UNRESOLVEDVulnerability lookup incomplete (scan --check-vulns)INFOMCP04
ATT-…Hash attestation finding (Layer 1 --verify-hashes). (ID generated at scan time)variesMCP04

Approved servers and drift 24 checks

What changed since you approved a setup: servers added, removed or launched differently, a hand-edited lock file, and, with --connect, tools added or changed since approval. A server started through mcp-audit run is checked against the lock as it launches and while it runs, and a policy can require every server to be wrapped that way.

Runs with mcp-audit lock --verify mcp-audit check mcp-audit wrap

IDWhat it flagsSeverityOWASP MCP
BL-001Baseline file could not be parsednot ratednone
LOCK-001Server drifted from lockHIGHMCP04
LOCK-002Server not in lockHIGHMCP04
LOCK-003Locked server missingMEDIUMMCP04
LOCK-004Resolution driftedCRITICAL or HIGHMCP04
LOCK-005Lock file tampered or hand-editedCRITICALMCP04
LOCK-006Tool added since lockHIGHMCP02 MCP04
LOCK-007Tool contract changed since lockHIGHMCP03 MCP04
LOCK-008Tool description changed since lockMEDIUMMCP03 MCP04
LOCK-009Tool removed since lockLOWMCP04
LOCK-010Resolution context changedINFOMCP04
LOCK-011Lock section digest does not holdHIGHMCP04
LOCK-012Lock file is not in canonical formLOWMCP04
LOCK-013Launch arguments carry a credentialLOWMCP04
LOCK-014Repository guardrail file loosened since the lockHIGHMCP02
RUGPULL-000First scan — baseline recorded for serverINFOnone
RUGPULL-001Server configuration changed since last scanHIGHMCP03 MCP04
RUGPULL-002New MCP server detected since last scanINFOMCP09
RUGPULL-003Previously tracked server no longer configuredINFOnone
RUN-001No lock covers this serverHIGHMCP04
RUN-002Lock unreadable at launchHIGHMCP04
RUN-003No tool record for a launched serverHIGHMCP04
WRAP-001Server not launched through mcp-audit runHIGHMCP04
DRIFT-…Baseline drift finding. (ID generated at scan time)variesMCP09

Community rules 41 checks

YAML rules in an open format that run on every scan: risky launch commands, inline code, shell metacharacters, sensitive variable names, sampling, and more.

Run on every scan. mcp-audit rule list shows the rules that are loaded.

40 of 41 community rules run by default; STDIO-001 ships disabled. The rule format is Apache 2.0 and open to contributions.

IDWhat it flagsSeverityOWASP MCP
COMM-001Prohibited network binary used as MCP serverHIGHMCP05
COMM-002curl used as MCP server binaryMEDIUMMCP05
COMM-003Browser sandbox disabledMEDIUMMCP05 MCP10
COMM-004Unrecognized stdio server binaryLOWMCP05 MCP09
COMM-005/tmp referenced in server argsMEDIUMMCP05
COMM-006Inline code execution via -c flagHIGHMCP05
COMM-007eval or exec passed in server argsHIGHMCP05
COMM-008SSH agent socket exposed to MCP serverMEDIUMMCP01
COMM-009AWS session token exposed to MCP serverHIGHMCP01
COMM-010npx used without pinned versionLOWMCP04
COMM-011Broad Deno permission flags in useMEDIUMMCP02
COMM-012Server binds to all interfaces (0.0.0.0)HIGHMCP07
COMM-013npx auto-confirm flag (silent execution)HIGHMCP04 MCP05
COMM-014MCP Sampling capability declaredLOWMCP04 MCP06
COMM-015Shell metacharacters in server argsCRITICALMCP05
COMM-016Role-claim poisoning in server argsHIGHMCP01
COMM-017Authority-claim poisoning — AI identity impersonationCRITICALMCP01
COMM-018Output-format hijack in server argsMEDIUMMCP01
COMM-019Side-channel data exfiltration instruction in server argsHIGHMCP01 MCP02
COMM-020Confabulation trigger in server argsHIGHMCP01
COMM-021Env key name suggests private or secret keyHIGHMCP02
COMM-022Env key name suggests access tokenMEDIUMMCP02
COMM-023Env key name suggests passwordHIGHMCP02
COMM-024HTTP endpoint on sensitive or production portMEDIUMMCP06
COMM-025Localhost endpoint with explicit SSE/HTTP transport URLLOWMCP06 MCP05
COMM-026Unencrypted WebSocket URL in server argsMEDIUMMCP06
COMM-027Generic or unnamed server keyLOWMCP07
COMM-028Command is an absolute pathMEDIUMMCP04
COMM-029Command references a temp or volatile directoryHIGHMCP04
COMM-030Generic-named server with network transport URLLOWMCP09 MCP07
COMM-031Remote MCP server configured without authenticationLOWMCP06
COMM-033Project-level MCP config file present in repositoryLOWMCP09
COMM-034Overprivileged MCP server credential (potential god key)MEDIUMMCP06
COMM-035TLS certificate checking disabled by environment variableMEDIUMMCP01
COMM-036Container launched with host takeover flagsHIGHMCP05
COMM-037Container launched on the host networkMEDIUMMCP05
COMM-038Package installed from an extra package indexMEDIUMMCP04
COMM-039Default package index replaced or custom npm registry configuredLOWMCP04
STDIO-001Launcher binary not in allowlist (ships disabled)MEDIUMMCP04 MCP09
STDIO-002aHidden (dotfile) launcher binaryHIGHMCP04 MCP05
STDIO-002bInline script launcherHIGHMCP05

More surfaces 4 checks

IDs generated at scan time for IDE extensions, MCP server source code, hash pins, baselines and your own governance policy.

Runs with mcp-audit extensions scan mcp-audit sast <path>

IDWhat it flagsSeverityOWASP MCP
EXT-…IDE extension security finding. (ID generated at scan time)variesMCP04
SAST-…SAST finding from Semgrep rule execution. (ID generated at scan time)variesMCP01 MCP05 MCP07
GOV-…Governance policy finding. (ID generated at scan time)variesnone
VULN-…Known-vulnerability finding from scan --check-vulns (network, opt-in). (ID generated at scan time)variesMCP04

OWASP MCP Top 10

Findings carry their OWASP MCP Top 10 codes in terminal, JSON and SARIF output. The mapping is machine-readable (OWASP 2025-beta).

MCP01 Token Mismanagement and Secret Exposure 38 checks

CFHYG-001, CFHYG-002, CFHYG-003, CFHYG-004, CFHYG-005, CFHYG-006, CFHYG-007, CFHYG-008, COLLIDE-001, COMM-008, COMM-009, COMM-016, COMM-017, COMM-018, COMM-019, COMM-020, COMM-035, CRED-001, CRED-002, CRED-003, CRED-004, CRED-005, CRED-006, MEM-006, MOD-002, POISON-001, POISON-002, POISON-003, POISON-041, REACH-001, REACH-002, REACH-003, SKILL-001, SKILL-006, TOXIC-003, TOXIC-009, WORKFLOW-001, WORKFLOW-002

MCP02 Privilege Escalation via Scope Creep 31 checks

AGENT-001, AGENT-002, COMM-011, COMM-019, COMM-021, COMM-022, COMM-023, GUARD-001, GUARD-002, HARNESS-001, HARNESS-003, HARNESS-004, HARNESS-005, HARNESS-006, HOOK-004, HOOK-005, LOCK-006, LOCK-014, MANAGED-004, MEM-008, PERM-001, PERM-003, PERM-005, PERM-007, SANDBOX-001, SANDBOX-002, SANDBOX-003, SANDBOX-004, SANDBOX-005, SKILL-006, TRANSPORT-002

MCP03 Tool Poisoning 44 checks

ANNOT-001, DELEG-001, DELEG-002, DELEG-003, INSTR-002, LOCK-007, LOCK-008, MEM-001, MEM-002, MEM-005, MEM-006, MEM-007, MEM-008, MOD-003, POISON-001, POISON-002, POISON-003, POISON-010, POISON-011, POISON-012, POISON-020, POISON-021, POISON-030, POISON-040, POISON-041, POISON-042, POISON-043, POISON-044, POISON-045, POISON-050, POISON-060, POISON-070, POISON-071, POISON-072, POISON-073, POISON-074, POISON-075, POISON-076, RUGPULL-001, SKILL-001, SKILL-002, SKILL-003, SKILL-004, SKILL-007

MCP04 Software Supply Chain Attacks 62 checks

ATTEST-011, ATTEST-012, ATTEST-013, ATTEST-014, COMM-010, COMM-013, COMM-014, COMM-028, COMM-029, COMM-038, COMM-039, EXTPOL-001, EXTPOL-002, EXTPOL-003, LOCK-001, LOCK-002, LOCK-003, LOCK-004, LOCK-005, LOCK-006, LOCK-007, LOCK-008, LOCK-009, LOCK-010, LOCK-011, LOCK-012, LOCK-013, MOD-001, MOD-002, MOD-004, MOD-005, MOD-006, PLUGIN-001, PLUGIN-002, PLUGIN-003, PLUGIN-004, RUGPULL-001, RUN-001, RUN-002, RUN-003, SC-001, SC-002, SC-003, SC-004, SC-005, SC-007, SC-008, SC-009, SC-010, SC-011, SKILL-003, SKILL-005, SKILL-006, SKILL-007, SKILL-008, STDIO-001, STDIO-002a, TRANSPORT-003, VULN-UNPINNED, VULN-UNRESOLVED, WORKFLOW-003, WRAP-001

MCP05 Command Injection and Execution 45 checks

AGENT-003, COMM-001, COMM-002, COMM-003, COMM-004, COMM-005, COMM-006, COMM-007, COMM-013, COMM-015, COMM-025, COMM-036, COMM-037, HARNESS-004, HARNESS-005, HARNESS-006, HOOK-001, HOOK-002, INTEG-001, MOD-004, PERM-003, PERM-004, PERM-006, POISON-074, SANDBOX-001, SANDBOX-002, SKILL-005, SKILL-006, SKILL-008, STDIO-002a, STDIO-002b, TOXIC-001, TOXIC-002, TOXIC-005, TOXIC-006, TOXIC-008, TOXIC-010, TOXIC-011, TRANSPORT-002, TRUST-002, TRUST-003, TRUST-006, TRUST-007, TRUST-008, WORKFLOW-003

MCP06 Intent Flow Subversion 28 checks

ANNOT-001, AUTH-001, AUTH-002, COMM-014, COMM-024, COMM-025, COMM-026, COMM-031, COMM-034, DELEG-001, DELEG-002, DELEG-003, MEM-004, MEM-005, MEM-007, MOD-003, POISON-010, POISON-011, POISON-012, POISON-030, POISON-070, POISON-071, POISON-072, POISON-073, POISON-075, POISON-076, WORKFLOW-001, WORKFLOW-002

MCP07 Insufficient Authentication and Authorization 25 checks

AGENT-001, AGENT-003, CFHYG-005, COMM-012, COMM-027, COMM-030, HARNESS-001, HARNESS-002, HARNESS-003, HARNESS-007, HOOK-001, HOOK-002, HOOK-004, HOOK-005, MANAGED-005, MANAGED-006, MOD-001, PERM-001, PERM-002, PERM-004, PERM-006, PERM-007, SKILL-006, TRANSPORT-001, TRANSPORT-004

MCP08 Lack of Audit and Telemetry 4 checks

CFHYG-006, CFHYG-007, CFHYG-008, MEM-004

MCP09 Shadow MCP Servers 24 checks

AGENT-004, CFHYG-002, COLLIDE-001, COMM-004, COMM-030, COMM-033, HARNESS-002, HARNESS-007, MANAGED-001, MANAGED-002, MANAGED-003, MANAGED-004, MANAGED-005, MANAGED-006, MANAGED-007, RUGPULL-002, SKILL-006, STDIO-001, TRUST-001, TRUST-002, TRUST-003, TRUST-004, TRUST-005, TRUST-008

MCP10 Context Injection and Over-sharing 21 checks

COMM-003, INSTR-001, INSTR-002, MEM-003, PERM-002, PERM-005, POISON-020, POISON-021, SANDBOX-003, SANDBOX-005, TOXIC-001, TOXIC-002, TOXIC-003, TOXIC-004, TOXIC-005, TOXIC-006, TOXIC-007, TOXIC-008, TOXIC-009, TOXIC-010, TOXIC-011

Supported clients 12 clients

mcp-audit finds these MCP config files on its own. Pass --path to scan any other file or folder.

ClientConfig file
Claude Desktop~/Library/Application Support/Claude/claude_desktop_config.json
Cursor~/.cursor/mcp.json
VS Code.vscode/mcp.json (workspace)
Windsurf~/.codeium/windsurf/mcp_config.json
Claude Code (user)~/.claude.json
Claude Code (project).mcp.json (project root)
GitHub Copilot CLI~/.copilot/mcp-config.json
Augment Code~/.augment/settings.json
Amazon Q Developer (project).amazonq/mcp.json (project root; scan --project only)
Codex$CODEX_HOME/config.toml (default ~/.codex/config.toml); .codex/config.toml (project) — TOML [mcp_servers.<name>]
Gemini CLI~/.gemini/settings.json; .gemini/settings.json (project) — mcpServers, comments allowed
OpenCode~/.config/opencode/opencode.json(c), $OPENCODE_CONFIG; opencode.json(c) (project) — mcp, JSONC

Files read in a repository 42 paths

mcp-audit scan --project <dir> walks the tree for these paths. The injection and obfuscation checks on instruction files need --include-agent-files; imports, heartbeat files, skill packages, subagents, git settings and workflows are checked in a default scan. Symlinked directories are reported, never followed.

PathWhat it isClient
.mcp.jsonMCP config or agent settings (servers start from it)claude-code
.claude/settings.jsonMCP config or agent settings (servers start from it)claude-code
.claude/settings.local.jsonMCP config or agent settings (servers start from it)claude-code
.cursor/mcp.jsonMCP config or agent settings (servers start from it)cursor
.cursor/settings.jsonMCP config or agent settings (servers start from it)cursor
.vscode/mcp.jsonMCP config or agent settings (servers start from it)vscode
.amazonq/mcp.jsonMCP config or agent settings (servers start from it)amazon-q
.codex/config.tomlMCP config or agent settings (servers start from it)codex
.gemini/settings.jsonMCP config or agent settings (servers start from it)gemini-cli
opencode.jsonMCP config or agent settings (servers start from it)opencode
opencode.jsoncMCP config or agent settings (servers start from it)opencode
.vscode/tasks.jsonEditor file that can run a command when the folder opensvscode-tasks
.vscode/settings.jsonEditor file that can run a command when the folder opensvscode-settings
.gemini/.envEditor file that can run a command when the folder opensgemini-env
.codex/hooks.jsonEditor file that can run a command when the folder openscodex-hooks
.claude/scheduled_tasks.jsonEditor file that can run a command when the folder opensclaude-scheduled-tasks
.claude/workflowsEditor file that can run a command when the folder opensclaude-workflows
.mcphub/servers.jsonEditor file that can run a command when the folder opensmcphub-config
.claude/commands/*.mdRead by the agent as instructionsclaude-code
.claude/skills/**/SKILL.mdRead by the agent as instructionsclaude-code
.agents/skills/**/SKILL.mdRead by the agent as instructionsagent-skills
.gemini/skills/**/SKILL.mdRead by the agent as instructionsgemini-cli
.cursor/skills/**/SKILL.mdRead by the agent as instructionscursor
.codex/skills/**/SKILL.mdRead by the agent as instructionscursor
.github/skills/**/SKILL.mdRead by the agent as instructionscopilot
.claude/CLAUDE.mdRead by the agent as instructionsclaude-code
CLAUDE.mdRead by the agent as instructionsclaude-code
.cursor/rules/*.mdcRead by the agent as instructionscursor
.github/copilot-instructions.mdRead by the agent as instructionscopilot
.github/instructions/*.instructions.mdRead by the agent as instructionscopilot
.github/prompts/*.prompt.mdRead by the agent as instructionscopilot
.claude/agents/**/*.mdRead by the agent as instructions
CLAUDE.local.mdRead by the agent as instructionsclaude-code
.claude/rules/**/*.mdRead by the agent as instructionsclaude-code
AGENTS.mdRead by the agent as instructionsagents-md
.claude/AGENTS.mdRead by the agent as instructionsclaude-code
AGENTS.override.mdRead by the agent as instructionscodex
SOUL.mdRead by the agent as instructionsopenclaw
USER.mdRead by the agent as instructionsopenclaw
MEMORY.mdRead by the agent as instructionsopenclaw
HEARTBEAT.mdRead by the agent as instructionsopenclaw
.github/workflows/*.ymlGitHub Actions workflow, read for AI agent steps that run on text a stranger can writegithub-actions