Chain Builder

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill is designed to ingest and execute externally defined "chain specifications" (Workflow 2), creating a surface for indirect prompt injection. A malicious specification could override intended behaviors or inject instructions into subsequent processing stages.
  • Ingestion points: Loading chain definitions from user-provided specifications or external templates (SKILL.md).
  • Boundary markers: Absent; the skill does not provide instructions for the agent to use delimiters or "ignore" instructions when processing external specifications.
  • Capability inventory: Sequential prompt execution, intermediate validation, and potential command execution as seen in the examples (SKILL.md).
  • Sanitization: Absent; the skill does not detail methods for sanitizing or validating the content of the prompt templates within a chain.
  • [COMMAND_EXECUTION]: The "Tested Code Generator" example workflow describes a step to run tests on agent-generated code ("Action: Run tests"). This represents a dynamic execution pattern where code produced in earlier stages is executed by the system, which could be exploited if the generation process is influenced by untrusted data.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 02:42 AM
Security Audit — agent-trust-hub — Chain Builder