Skills
Navigate
skills/ruvnet/ruflo/agent-sona-learning-optimizer/Gen Agent Trust Hub

agent-sona-learning-optimizer

Warn

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill utilizes npx to download and execute the claude-flow@alpha package during runtime. This represents the installation of a dependency from an unknown source that is not identified on the trusted vendor list.
  • [COMMAND_EXECUTION]: The skill provides shell hooks to be run via npx during pre-task and post-task phases, executing command-line instructions to interface with the learning optimizer.
  • [INDIRECT_PROMPT_INJECTION]: The skill interpolates external task data into a shell command, creating a surface for injection attacks.
  • Ingestion points: Untrusted data enters the agent context via the $TASK variable in SKILL.md.
  • Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands within the $TASK string.
  • Capability inventory: Shell execution capability via the npx utility in SKILL.md.
  • Sanitization: Absent; the skill does not show evidence of escaping or validating the input before it is passed to the command line.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 15, 2026, 11:38 AM
Security Audit — agent-trust-hub — agent-sona-learning-optimizer