The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill transforms untrusted design input into structured implementation plans that include executable content.\n- Ingestion points: Design documents and architectural context provided as input to the planning process.\n- Boundary markers: Absent; the skill lacks delimiters or specific instructions to ignore malicious directives in the source data.\n- Capability inventory: Generates implementation plans that prescribe the execution of Python code and shell commands such as git and pytest.\n- Sanitization: Absent; the skill relies on the agent's internal safety filters rather than explicit validation.

Writing Plans