ln-814-optimization-executor

Pass

Audited by Gen Agent Trust Hub on May 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches reference files and configuration from the author's official GitHub repository (levnikolaevich/claude-code-skills) if local files are missing. This behavior is documented as a fallback mechanism for the skill's dependencies.
  • [COMMAND_EXECUTION]: Executes shell commands defined in local context files (test_command, e2e_test_command) to run benchmarks and safety tests. This is a core part of the optimization workflow.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads optimization hypotheses and execution parameters from project-level context files without explicit sanitization.
  • Ingestion points: Reads instructions and commands from .hex-skills/optimization/{slug}/context.md using the mcp__hex-line__read_file tool.
  • Boundary markers: None; the data is read directly into the agent's context.
  • Capability inventory: Full shell access (Bash, Monitor), git operations, and multi-file editing capabilities.
  • Sanitization: None; the skill relies on the provided commands and hypotheses to drive the optimization process.
Audit Metadata
Risk Level
SAFE
Analyzed
May 12, 2026, 06:29 AM