research-planner

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is a conduit for indirect prompt injection because it processes untrusted user input into a structured file that downstream skills (e.g., a task-splitter) are instructed to act upon.\n
  • Ingestion points: The research topic and task description provided by the user in Step 1 and Step 2 are interpolated directly into the instructions and the YAML frontmatter of the output file.\n
  • Boundary markers: The skill uses YAML frontmatter for metadata but does not wrap the user-influenced instructions in delimiters or provide 'ignore instructions' warnings to the downstream skills that will read the file.\n
  • Capability inventory: The skill possesses file-write capabilities, creating directories and Markdown files at research/<slug>/<slug>-research-plan.md.\n
  • Sanitization: The skill implements a slugification routine for the filename (lowercase, alphanumeric replacement, hyphen collapsing), which effectively prevents path traversal. However, it lacks sanitization of the research instructions' content.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 01:28 PM