tw-research-stat-consultant

Pass

Audited by Gen Agent Trust Hub on May 1, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an architectural surface for indirect prompt injection because it ingests and processes untrusted data from user-uploaded files (CSV, Excel, SPSS) and external reference documents.
  • Ingestion points: Data enters the agent's context through user-provided files like your_data.csv (via pd.read_csv) and local repository files such as ../../tw_edu_concept_alignment.md.
  • Boundary markers: The skill lacks explicit delimiters or instructions (e.g., XML tags or clear 'ignore instructions in this file' warnings) to prevent the agent from interpreting content within these data files as executable instructions or behavioral overrides.
  • Capability inventory: The agent is equipped with powerful tools, including Bash for shell command execution, Write for file modification, and WebSearch for network access.
  • Sanitization: No sanitization or verification steps are defined to filter potential malicious payloads that might be embedded in the structured or unstructured data being analyzed.
Audit Metadata
Risk Level
SAFE
Analyzed
May 1, 2026, 12:51 PM