The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data by taking user input through the $1 parameter and instructing the agent to investigate the existing codebase. This creates an attack surface where malicious instructions embedded in files or user prompts could potentially influence the agent's output.
Ingestion points: The $1 argument and local codebase files targeted during the investigation phase.
Boundary markers: Absent; the skill does not use specific delimiters or instructions to treat the codebase content as untrusted data.
Capability inventory: The skill utilizes file system investigation and the AskUserQuestion tool for interaction.
Sanitization: Absent; there is no explicit logic to filter or escape content retrieved from the codebase.
[SAFE]: The skill is purely instructional and does not include any executable scripts, network exfiltration attempts, or hardcoded credentials. It explicitly instructs the agent to avoid technical implementation details (architecture, APIs, libraries), which aligns with best practices for early-stage requirement gathering.

create-requirements