The Agent Skills Directory

[SAFE]: The skill enforces multiple security-oriented operational rules, such as requiring backup branches before history-altering commands (rebase, reset, amend) and mandating the use of --force-with-lease with specific SHAs for push operations to prevent accidental data loss or overwriting of remote history.
[COMMAND_EXECUTION]: The bisect_oracle.py script executes test commands provided by the agent during a git bisect workflow. The implementation uses the subprocess.run API with list-based arguments and shell=False, which prevents shell injection vulnerabilities. The instructions further mitigate risk by requiring the agent to use absolute paths for test scripts stored outside the repository being analyzed.
[EXTERNAL_DOWNLOADS]: The skill recommends the installation of standard development tools such as commitlint and lefthook for repository maintenance. The instructions specify using the project's native package manager and only performing these actions if requested by the user, adhering to standard development practices for well-known tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from git logs and status outputs. While this is an ingestion surface, the risk is mitigated by instructions that require the agent to follow established project styles and seek user confirmation before replacing existing configurations or performing significant history rewrites.

commit-and-rewrite-git