setup

Warn

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill instructs the agent to read sensitive credentials from the local file ~/.config/catalyst-cloud/customer.json and transmit them via a curl request to a remote endpoint. While it advises the agent not to print the key in the terminal, the credential is still exposed to the network destination specified in the configuration.
  • [EXTERNAL_DOWNLOADS]: The checklist includes the execution of npx @catalyst-cloud/catalyst-skills, which downloads and runs external code from the npm registry at runtime. The package is not associated with a recognized trusted organization.
  • [REMOTE_CODE_EXECUTION]: The use of npx to execute a package from a public registry constitutes remote code execution, as the package content is fetched and executed on the local machine during the readiness check.
  • [COMMAND_EXECUTION]: The skill relies on executing several shell commands, including curl for network requests, npx for package invocation, and ls for filesystem inspection.
  • [INDIRECT_PROMPT_INJECTION]: Vulnerability surface identified. 1. Ingestion point: ~/.config/catalyst-cloud/customer.json. 2. Boundary markers: Absent. 3. Capability inventory: Network requests via curl, package execution via npx, and file system inspection via ls. 4. Sanitization: Absent. Maliciously crafted data within the local configuration file could be used to manipulate the agent's behavior during the execution of these capabilities.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 10, 2026, 08:08 AM
Security Audit — agent-trust-hub — setup