prompt-enhance
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a Python script via a shell command in Step 2:
python3 {PROMPT_ENGINE_DIR}/scripts/search_prompts.py "KEY_TERMS" --limit 5. This pattern allows the agent to run local code based on dynamic parameters. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user prompts to generate the
KEY_TERMSused in the shell command described above. This creates a potential attack surface for command injection. - Ingestion points: User-provided prompts are processed in 'Step 1: Analyze the Input Prompt' and then used to generate terms for 'Step 2: Find Similar Top Prompts'.
- Boundary markers: The instructions do not define delimiters (like XML tags or triple quotes) or provide specific instructions to the agent to treat the input solely as data.
- Capability inventory: The skill uses
python3shell execution to query its prompt database. - Sanitization: There are no explicit instructions to sanitize or escape the
KEY_TERMSbefore interpolation into the shell command, which may lead the agent to include malicious characters (like semicolons or backticks) if present in the user prompt.
Audit Metadata