write-skill

Pass

Audited by Gen Agent Trust Hub on Nov 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONNO_CODE
Full Analysis
  • [COMMAND_EXECUTION]: Several Python and Node.js scripts in the scripts/ and root directories use subprocess calls (subprocess.run, subprocess.Popen, execSync) to orchestrate the evaluation workflow.
  • scripts/run_eval.py and scripts/improve_description.py execute the claude CLI with the -p flag to run automated evaluation queries and description improvements.
  • render-graphs.js executes the system dot command (from Graphviz) to convert markdown code blocks into SVG diagrams.
  • [DYNAMIC_EXECUTION]: The skill implements an evaluation loop in scripts/run_loop.py that dynamically generates improved skill descriptions based on LLM feedback and then re-evaluates them using automated tests.
  • [NO_CODE]: While the skill contains many utility scripts for the developer, the core SKILL.md is primarily instructional markdown and YAML frontmatter meant to guide an agent's behavior during skill authoring.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 7, 2026, 09:12 AM