jmb-wizard

Pass

Audited by Gen Agent Trust Hub on Nov 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository files to determine its configuration logic.
  • Ingestion points: SKILL.md instructs the agent to read project files like .env, README, docker-compose, and .github/workflows/* to scope the wizard.
  • Boundary markers: No explicit delimiters are used to wrap or sanitize the ingested content within the prompt context.
  • Capability inventory: The skill produces bash scripts that utilize the gh CLI for secret management and perform local file operations.
  • Sanitization: No sanitization or validation of the ingested repository content is performed.
  • [DYNAMIC_EXECUTION]: The skill dynamically generates a bash script by populating a template with logic derived from the repository structure.
  • [COMMAND_EXECUTION]: The agent performs shell operations including syntax checking (bash -n) and permission changes (chmod +x) on the generated files. The resulting wizard script executes various shell utilities like grep, mktemp, and the GitHub CLI.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 19, 2026, 07:02 PM