Salesloft
Analysis
This appears to be a real Salesloft integration, but it asks the agent to install an unpinned global CLI and gives broad delegated Salesloft access without clear approval, scope, or credential-boundary limits.
Findings (8)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.
`clientAction.agentInstructions` (optional) — instructions for the AI agent on how to proceed programmatically.
The skill allows external connection responses to provide agent-facing instructions. This is purpose-aligned for setup, but the agent should treat those instructions as lower priority than user and system instructions.
Popular actions ... `Create Person | create-person` ... `When the available actions don't cover your use case, you can send requests directly to the Salesloft API through Membrane's proxy.`
The skill exposes both action execution and a direct API proxy for Salesloft, including mutating record operations, without explicit approval gates, safe-operation limits, or rollback guidance.
`npm install -g @membranehq/cli@latest`
The skill instructs installation of an unpinned latest-version npm package globally, despite having no install spec. This creates a package provenance and version-drift risk.
`npx @membranehq/cli connection get <id> --wait --json`
The skill directs the agent to execute npm-hosted CLI code at runtime. That is not inherently malicious, but it is unexpected for an instruction-only skill with no declared required binaries or install spec.
Manage sales data, records, and workflows ... Cadence Enrollment ... Do Not Contact List ... Create Person
The skill can interact with shared sales records and workflow-related objects. A mistaken or overbroad action could propagate through team sales processes without containment guidance.
Checks whether tool use, credentials, dependencies, identity, account access, or inter-agent boundaries are broader than the stated purpose.
Primary credential: none ... Capability signals: `requires-oauth-token`, `requires-sensitive-credentials`
The declared credential contract says no primary credential, but the capability signals indicate OAuth and sensitive credential use, creating an under-disclosed permission boundary.
Checks for exposed credentials, poisoned memory or context, unclear communication boundaries, or sensitive data that could leave the user's control.
`clientAction.agentInstructions` (optional) — instructions for the AI agent on how to proceed programmatically.
External setup responses can add instructions into the agent's context. This is useful for integration setup but should not be treated as trusted memory or higher-priority policy.
This skill uses the Membrane CLI to interact with Salesloft. Membrane handles authentication and credentials refresh automatically.
Salesloft access is mediated by a third-party provider/gateway that manages authentication and refresh. This is purpose-aligned but important for users to understand as a data and credential boundary.
