The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/prompt_tester.py script enables the execution of external commands through the run_runner function. The script constructs a shell command by formatting a user-provided template string (--runner-cmd) with prompt and input data. This allows for the execution of arbitrary system commands if the runner configuration or the input data is maliciously crafted. Evidence in scripts/prompt_tester.py: The code uses cmd = runner_cmd.format(prompt=prompt, input=case_input) followed by subprocess.run(shlex.split(cmd), ...).
[DYNAMIC_EXECUTION]: The skill facilitates the dynamic assembly and execution of shell commands at runtime based on external parameters. The command structure is determined by values passed to the script, which may include untrusted content from test cases or prompt files.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes external prompts and test case data while maintaining shell execution capabilities. 1. Ingestion points: scripts/prompt_tester.py and scripts/prompt_versioner.py ingest data from local files, stdin, and CLI arguments. 2. Boundary markers: Absent. The skill does not implement specific delimiters or instructions to the agent to ignore embedded instructions in the test cases or prompts. 3. Capability inventory: scripts/prompt_tester.py contains the capability to execute shell commands using subprocess.run. 4. Sanitization: Absent. The script lacks sanitization or escaping mechanisms for the prompt and input variables before they are interpolated into the shell command template.

prompt-engineer-toolkit