The Agent Skills Directory

[SAFE]: The skill implements significant safety guardrails for potentially destructive Git operations. It mandates the creation of recovery points (backups or reflog notes) before any history-rewriting step. It also strictly requires the use of --force-with-lease for push operations to prevent overwriting remote work unknowingly. The logic relies on observed command output rather than agent assumptions, maintaining a clear separation between planning and execution.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a natural ingestion surface for untrusted data when processing Git conflicts and history, though this is inherent to its primary purpose.
Ingestion points: references/git-repair-method.md specifies reading external file content and differences via git diff --merge and git log --merge -p -- <file>.
Boundary markers: Absent. The instructions do not explicitly define delimiters to separate file content from agent instructions, nor do they include warnings to ignore instructions embedded within files.
Capability inventory: The skill generates plans involving powerful Git commands such as git push --force-with-lease, git rebase, and git reset --hard.
Sanitization: Absent. No specific content filtering or sanitization is mentioned for the text being resolved during a merge.
Mitigation: The risk is mitigated by the skill's design, which emphasizes human-in-the-loop plan execution and the reporting of observed outputs only.

omh-git-workflow