The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md direct the agent to execute a local Python script scripts/budget_check.py to validate user-supplied budget configurations. This script performs arithmetic validation using only Python standard libraries.
[INDIRECT_PROMPT_INJECTION]: The skill processes external JSON data provided by the user via the budget_check.py script. The script's output is ingested by the agent to report on feasibility, creating an inherent attack surface where malicious content in the JSON could attempt to influence agent behavior, though the script itself does not execute data as code.
Ingestion points: User-provided JSON file path passed to scripts/budget_check.py.
Boundary markers: None explicitly defined in the script output for the agent.
Capability inventory: Local script execution (python3) and file system read/write for template generation.
Sanitization: The script uses json.loads() and strict type checking (integers and finite floats) to validate the input structure.
[EXTERNAL_DOWNLOADS]: The documentation in references/hardware-budgets.md contains reference links to external technical resources, including the ares-emulator source code on GitHub and technical development blogs (plutiedev.com), which are used for hardware corroboration.

megadev-planning