backlog

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted feature descriptions from a backlog file, which are then used to generate implementation plans. This creates a surface for indirect prompt injection.\n
  • Ingestion points: The triager agent reads task descriptions from .backlog/backlog.md and populates PRD files (triage/SKILL.md).\n
  • Boundary markers: Absent. The instructions do not specify any delimiters or safety prompts to isolate user-provided task descriptions from the agent's instructions.\n
  • Capability inventory: The implementer agent (implement/SKILL.md) has the capability to modify source code and execute git commands as well as project-specific linting and testing tools.\n
  • Sanitization: The workflow relies on manual human verification of the 'approved' status in the PRD and Plan files before the implementer agent acts.\n- [COMMAND_EXECUTION]: The implementer agent is instructed to execute local system commands to manage the development branch and verify changes.\n
  • Evidence: In implement/SKILL.md, the agent is directed to use git pull, git checkout, git rebase, and git commit. It is also instructed to 'Run relevant lint/tests'.\n
  • Mitigation: The skill includes explicit rules against git push, force pushing, and automatic PR creation, which limits the scope of command execution to the local environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 01:16 PM