burndown-full
Warn
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill directs the agent to autonomously discover and execute verification commands such as build, test, lint, and typecheck scripts defined within the repository's configuration files (e.g., package.json, Makefile, justfile, or README). This creates a risk of executing malicious commands if the project environment or repository contains compromised build scripts.
- [PROMPT_INJECTION]: The instructions include directives to override standard agent verification cycles and user oversight. Specifically, the skill mandates that the agent 'Never stop early to ask permission to continue' and 'Continue automatically until the worklist is exhausted,' which promotes high-autonomy execution of shell commands without human-in-the-loop review.
- [COMMAND_EXECUTION]: The skill exhibits a significant vulnerability surface for indirect injection via project metadata.
- Ingestion points: package.json, Makefile, justfile, README, AGENTS.md, CLAUDE.md, and CI configuration files.
- Boundary markers: Absent.
- Capability inventory: Full shell command execution capability for build, test, and lint cycles.
- Sanitization: None; the agent is instructed to note and run the exact commands found in these files.
Audit Metadata