prompt-engineer-toolkit

Pass

Audited by Gen Agent Trust Hub on Apr 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONNO_CODE
Full Analysis
  • [COMMAND_EXECUTION]: The skill documentation instructs the agent to execute local Python scripts (scripts/prompt_tester.py and scripts/prompt_versioner.py) and allows the use of a customizable runner command (--runner-cmd) which executes external CLI tools via shell commands.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it ingests and processes untrusted data from multiple sources.\n
  • Ingestion points: Data is read from testcases.json, prompts/a.txt, prompts/b.txt, and user-provided JSON payloads via the --input flag.\n
  • Boundary markers: There are no explicit instructions or delimiters mentioned in the skill to prevent the agent or the underlying scripts from obeying instructions embedded within the test cases or prompts.\n
  • Capability inventory: The skill has the capability to execute shell commands through the prompt_tester.py script by interpolating inputs into the --runner-cmd argument.\n
  • Sanitization: No sanitization, validation, or escaping of the input data is described, potentially allowing malicious inputs to influence command execution or model output.\n- [NO_CODE]: The core functional scripts referenced in SKILL.md (scripts/prompt_tester.py and scripts/prompt_versioner.py) are not present in the provided skill files, which prevents verification of their internal security logic and sanitization practices.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 5, 2026, 08:06 PM
Security Audit — agent-trust-hub — prompt-engineer-toolkit