The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands such as ls, grep, mkdir, and mktemp to interact with the project's filesystem and manage session data. This behavior is documented and aligns with the skill's purpose of researching and organizing product design documents.\n- [COMMAND_EXECUTION]: The skill includes an optional phase to execute an external CLI tool named codex for model-based second opinions. It verifies tool availability and executes the command in a read-only mode to minimize risk.\n- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests user-provided product ideas and keywords that are subsequently used in shell commands and sub-agent prompts.\n
Ingestion points: User-supplied text captured through interaction prompts in Phase 1 and Phase 2.\n
Boundary markers: The workflow uses distinct phases and specific task descriptions to provide context to the agent, though explicit delimiters for user data are not present.\n
Capability inventory: The skill has capabilities for filesystem read/write and external CLI execution.\n
Sanitization: No explicit sanitization or escaping of user-provided strings is performed before interpolation into shell command templates.

office-hours