limitr-ts

Fail

Audited by Snyk on Nov 21, 2026

Risk Level: CRITICAL
Full Analysis

CRITICAL E004: Prompt injection detected in skill instructions.

  • Potential prompt injection detected (medium risk: 0.30). The skill is a legitimate Limitr SDK integration guide, but its default for Cloud policies (httpHosts: true) lets the policy reach any public host over HTTP, which weakens the sandbox the doc describes for Limitr.new beyond what the advertised metering task needs.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (low risk: 0.10). The required Limitr code only ingests first-party policy, customer records, and usage counts, but the Claude Agent SDK recipe enables a WebSearch tool whose results are arbitrary outsider-authored web pages the agent must actively fetch and read at runtime.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

  • Direct money access detected (medium risk: 0.30). The skill is mainly a usage-metering and entitlement library for TypeScript. Most of it is excluded: gating, metering, reserve/settle of LLM usage, spend caps, and event handlers are internal usage governance, and the billing handler example is only illustrative. One documented route is operational and financial. On Limitr Cloud, applyCustomerTopup "asks Cloud to buy the topup: Cloud charges for it first, then the grant arrives" (lines 275-277). This is a purchase that charges a customer, so it is a direct operation (a charge/purchase). The Cloud route does require a token, Limitr.cloud({ token: process.env.LIMITR_TOKEN! }) (line 334). It is an authenticated session with validated service credentials, enforced by Cloud. Client tokens are scoped and "can't create or change customers" (lines 349-354). That is a documented service-side authentication mechanism for the route, so it counts as qualifying control. The Network module (network.connect(url, plan, spend, expires)) also involves spend, but its details are delegated to another skill, and it too runs over an authenticated Cloud connection. Because the direct operation has qualifying service authentication, the tier is medium rather than high.

MEDIUM W013: Attempt to modify system services in skill instructions.

  • Attempt to modify system services in skill instructions detected (low risk: 0.10). The skill exposes runtime operations that change the live policy engine (setPlan, deletePlan, setCapabilities, allowHttp, docCall) and customer state, but no explicit text establishes unrecoverable loss of operational state, so the answer stays at low.

Issues (4)

E004
CRITICAL

Prompt injection detected in skill instructions.

W011
MEDIUM

Third-party content exposure detected (indirect prompt injection risk).

W009
MEDIUM

Direct money access capability detected (payment gateways, crypto, banking).

W013
MEDIUM

Attempt to modify system services in skill instructions.

Audit Metadata
Risk Level
CRITICAL
Analyzed
Nov 21, 2026, 02:37 AM
Issues
4