token-burner

Warn

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes various system commands for project scanning and verification, including git, npm audit, govulncheck, and jq. It autonomously modifies the filesystem by creating git worktrees and executing tests on newly generated code.
  • [REMOTE_CODE_EXECUTION]: The skill recommends installing third-party tools such as pip-audit and suggests running complexity-report via npx, which involves downloading and executing packages from external registries.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted project data (source code, diffs, and comments) to generate task prompts.
  • Ingestion points: Files like package.json, go.mod, git diff output, and source code are read in SKILL.md (Step 1) and templates/task-discovery.md.
  • Boundary markers: Absent. The skill does not implement explicit delimiters or instructions to ignore embedded commands within the analyzed project content.
  • Capability inventory: Includes filesystem writes, subprocess execution (git, npm, go), and the ability to trigger an 'Agent' tool with generated prompts.
  • Sanitization: None described. Project content is directly used to influence the autonomous task discovery and prompt generation process.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 16, 2026, 11:46 AM
Security Audit — agent-trust-hub — token-burner