remote-run-curl

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is designed to construct and execute curl commands that send shell instructions to a remote runner managed by BuildBuddy. The steps field in the API request allows for arbitrary command execution on the remote infrastructure.
  • [INDIRECT_PROMPT_INJECTION]: The skill uses persistent-fields.md to store and retrieve configuration parameters that are automatically applied to future API requests. This creates a stateful injection surface where malicious instructions or flags could be persisted across sessions.
  • Ingestion points: The agent reads configuration from persistent-fields.md in the Field persistence section of SKILL.md.
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious embedded content within the persisted fields.
  • Capability inventory: The skill uses curl to perform network operations and executes remote commands via the BuildBuddy Run API.
  • Sanitization: The instructions include a requirement for the agent to ask for user confirmation before writing to the persistence file, which serves as a manual checkpoint but does not programmatically sanitize the content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 07:16 PM
Security Audit — agent-trust-hub — remote-run-curl