The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the uv tool to run Python scripts. This includes inline Python snippets using python -c and a local script scripts/sympy_compute.py managed by runtime.harness. While these are functional for math problem-solving, they involve direct interaction with the host shell.
[PROMPT_INJECTION]: An indirect prompt injection surface is present in the Sympy_Integrate tool command. The command structure uv run python -m runtime.harness scripts/sympy_compute.py integrate "{expression}" interpolates user-provided mathematical functions into a shell command.
Ingestion points: User-provided expressions for integration (e.g., "f(x)") in SKILL.md.
Boundary markers: Expressions are enclosed in double quotes, which can be bypassed if the input contains shell metacharacters.
Capability inventory: The skill has access to the Bash tool and uv execution environment.
Sanitization: No explicit sanitization or validation of the math expression is mentioned in the instructions, relying on the underlying agent or script to handle escaping.

numerical-integration