kill-switch-controls

Pass

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [SAFE]: No malicious patterns such as obfuscation, hidden instructions, or persistence mechanisms were found in any of the skill files. The technical content matches the stated purpose of financial research.
  • [COMMAND_EXECUTION]: The skill executes a local Python script (scripts/kill_switch_controls_diagnostics.py) to process financial data. The script is deterministic and uses the well-known pandas library. It requires specific CSV columns to be present, which serves as a validation step against processing unrelated files.
  • [PROMPT_INJECTION]: The skill processes external data via a CSV file, which presents an indirect prompt injection surface. Ingestion points: Data enters the process through scripts/kill_switch_controls_diagnostics.py reading input.csv. Boundary markers: None are explicitly used in the instructions to separate untrusted data. Capability inventory: The skill can execute Python code via subprocess and write files to disk. Sanitization: The script validates column headers and enforces numeric data types on metrics fields, which effectively prevents the interpretation of those fields as instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 17, 2026, 07:03 PM
Security Audit — agent-trust-hub — kill-switch-controls