The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/init_instruction.py performs file system operations to create new instruction files.
Evidence: The script uses os.path.join(output_dir, name) and open(path, 'w').write() to write content to the disk.
Risk: There is no validation or sanitization of the name or output arguments. This could theoretically allow a user to write files to unintended locations via path traversal (e.g., using ../), although the script enforces a .instructions.md file extension.
[PROMPT_INJECTION]: The skill is designed to create 'Policy Maker' files that define the behavior and constraints of AI agents (GitHub Copilot).
Ingestion points: User-provided inputs for description and applyTo are interpolated directly into the generated markdown frontmatter in scripts/init_instruction.py.
Boundary markers: The generated template lacks explicit boundary markers or instructions to the agent to ignore potentially malicious content within the interpolated fields.
Capability inventory: The generated instructions are intended to influence the agent's file processing and code generation logic.
Sanitization: No sanitization is performed on the input strings before they are written to the instruction files, creating a surface for indirect prompt injection if the generated files are shared or used in untrusted environments.

instruction-creator