create-backlog-item

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface in its autonomous mode where it processes untrusted data from local project files.
  • Ingestion points: Files located within the research/ directory and existing backlog items in ~/.dh/projects/{slug}/backlog/ are read to derive item fields (SKILL.md).
  • Boundary markers: The skill does not define structural boundary markers for the ingested content, relying on natural language instructions to separate data from instructions (SKILL.md).
  • Capability inventory: The skill uses the mcp__plugin_dh_backlog__backlog_add tool, which can write to the local filesystem and interact with the GitHub API to create issues (SKILL.md).
  • Sanitization: The instructions include a mandatory filtering step to strip implementation steps, suggested fixes, and file-level prescriptions from the derived description to prevent the agent from adopting solutions found in research data (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 08:40 AM