The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses simple shell commands in its pre and post hooks (echo, tail, date). These are used for status reporting and generating timestamps for internal memory storage. No dangerous commands or user-input injection into the shell were detected.
[DATA_EXFILTRATION]: The skill uses memory_store and memory_search to interact with the agent's internal memory state. There are no network operations (like curl or wget) or attempts to access sensitive system files.
[PROMPT_INJECTION]: The instructions provide architectural guidance for pseudocode generation. There are no attempts to override system safety guidelines, bypass constraints, or extract system prompts.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to process specifications (which may come from external sources), it lacks dangerous capabilities like file writing, network access, or arbitrary code execution. The risk of data-driven attacks is negligible given the restricted toolset.

agent-pseudocode