The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing user-supplied prompts and context through an LLM for optimization. This content is ingested without boundary markers or sanitization, creating an attack surface where malicious instructions embedded in the input data could influence the optimization logic.
Ingestion points: The scripts/optimize_prompt.py script reads external content from paths provided to the --prompt-file and --context-file arguments.
Boundary markers: The script does not utilize delimiters or specific instructions to isolate the untrusted content from the optimizer's system instructions.
Capability inventory: The skill possesses capabilities for reading local files, writing optimized prompts to disk, and performing network operations via the Copilot runtime API.
Sanitization: No validation or filtering is performed on the ingested content before it is passed to the language model.- [COMMAND_EXECUTION]: The skill includes a Python script (scripts/optimize_prompt.py) intended for command-line execution. This script handles file operations and network requests to facilitate prompt optimization workflows.- [DYNAMIC_EXECUTION]: The optimization script utilizes the DSPy library to compile and update prompt optimization programs at runtime. These programs are data-driven configurations that govern how the LLM optimizes prompts, representing a form of runtime logic adjustment based on input data.- [EXTERNAL_DOWNLOADS]: The skill relies on external Python dependencies, specifically dspy and pyyaml. The documentation directs users to install these standard libraries from public package registries to enable the optimization functionality.

engineer-prompt