The Agent Skills Directory

[SAFE]: Comprehensive analysis of the provided instructions, specifications, and scripts shows no evidence of malicious intent, data exfiltration, or unauthorized persistence.
[COMMAND_EXECUTION]: The helper script scripts/ulysses.sh performs local state management using standard development commands such as git add, git commit, and jq filters.
[REMOTE_CODE_EXECUTION]: The skill utilizes the thoughtbox_execute tool to run agent-authored validation code (validator cells) within a cloud-based environment, which is the core intended mechanism of the protocol.
[PROMPT_INJECTION]: The protocol includes built-in mitigations for indirect prompt injection by requiring the agent to define and 'freeze' its validation logic before it observes the actual system data (the 'observed' JSON), preventing the data from manipulating the verification process.

ulysses-protocol