task-decomposer

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its method of ingesting and acting upon external data.\n
  • Ingestion points: External data enters through user task descriptions and domain-specific configuration files located at skills/.curated/domains/<domain>/ECOSYSTEM.md.\n
  • Boundary markers: Absent; the instructions do not implement markers or warnings to disregard instructions within the ingested content.\n
  • Capability inventory: High; the skill has permissions to write code, create files, and execute tasks.\n
  • Sanitization: Absent; no validation or sanitization of input data or domain paths is defined.\n- [COMMAND_EXECUTION]: The skill is authorized to perform file system operations and code generation during task execution, which could be leveraged if malicious instructions are successfully injected.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 06:36 AM
Security Audit — agent-trust-hub — task-decomposer