The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided specifications or requirements (untrusted data) to generate implementation plans. These plans include structured tasks with code snippets and shell commands. If a user provides malicious instructions within their specification, the agent may inadvertently include them in the generated plan.
Ingestion points: User specifications and requirements provided at runtime (SKILL.md).
Boundary markers: Absent; the skill uses standard Markdown headers for the plan but does not provide specific instructions to the agent to ignore embedded instructions in the source requirements.
Capability inventory: The generated plans involve creating/modifying files and executing shell commands like pytest and git (SKILL.md).
Sanitization: Absent; the skill relies on the agent's default logic to process the input.
[DYNAMIC_EXECUTION]: The skill generates Python code and Bash scripts from templates to be executed by the agent or a subagent during the implementation phase. This script generation is a core function of the skill's planning purpose.

writing-plans