The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of Git commands (git diff, git status, git commit, git push) to perform its primary function. These are governed by strict rules, including mandatory human approval of the exact commit messages and files before any modification or network operation occurs.
[EXTERNAL_DOWNLOADS]: The protocol.yon file includes a reference to an external guide at https://yon.younndai.com/yon-guide.txt. This is documented as a reference link and does not involve automated script execution or package installation.
[DATA_EXFILTRATION]: Although the skill transmits data to a remote repository via git push, it implements robust safeguards to prevent accidental exfiltration of sensitive information. It explicitly instructs the agent to exclude .env files, credentials, binaries, and debug artifacts from the sync process.
[PROMPT_INJECTION]: The skill processes untrusted data from git diff outputs, creating a surface for indirect prompt injection. This is mitigated by the 'Step 4' approval gate, where a human must review the proposed commit plan and messages before the agent takes any action.
[SAFE]: The skill exhibits high security maturity by implementing 'Self-improvement' through a controlled protocol that requires user-approved diffs, rather than autonomous code modification.

github-sync