record-episode
Warn
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using
python3 harness/scripts/mem-rank.pyin the 'Steps' section. - [COMMAND_EXECUTION]: Multiple user-provided variables, including
<did>,<kết quả>,<tên/issue>, and<câu hỏi>, are interpolated directly into shell command templates. This creates a command injection vulnerability where malicious input containing shell metacharacters could be used to execute arbitrary code on the underlying system. - [PROMPT_INJECTION]: The skill ingests untrusted data (session descriptions and results) and incorporates it into sensitive execution contexts without sanitization.
- Ingestion points: Arguments provided in the shell command templates in
SKILL.md. - Boundary markers: None identified in the command construction.
- Capability inventory: Execution of local Python scripts via the shell.
- Sanitization: No sanitization or input validation logic is present in the skill instructions.
Audit Metadata