icp-sequencer-workflow
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input such as buyer roles, company criteria, and outreach copy which are then used to construct search queries and campaign content within the Clay platform.
- Ingestion points: Input fields for 'Buyer roles and seniority', 'Company criteria', and 'Seller, offer and CTA' in SKILL.md.
- Boundary markers: The skill does not provide explicit delimiters or instructions to the agent to treat user-provided ICP criteria as untrusted or to ignore embedded instructions during interpolation.
- Capability inventory: The skill creates automated workflows with search triggers, writes records to Audiences, and manages outbound campaigns via CLI tools.
- Sanitization: No sanitization or validation steps are defined for the user-supplied data before it is used to generate the dslQuery or campaign outreach.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the 'clay' CLI utility to perform workspace discovery and manage platform resources.
- Evidence: Instructions include running
clay whoami,clay workflows nodes --help, and other commands for creating and reading back artifacts in sections 'Discovery and artifact links' and '2. Build the native workflow'.
Audit Metadata