The Agent Skills Directory

[SAFE]: The skill implements a legitimate workflow for Git conflict resolution using established tools and bundled scripts. It includes specific safety checks, such as detecting squash-merge residue and requiring user confirmation before applying remedies.
[COMMAND_EXECUTION]: The skill executes standard Git commands (git log, status, config, rerere, mergetool) and a bundled Python ZipApp (melt.pyz) located within the skill's own directory. These executions are scoped to the intended purpose of conflict resolution and repository management.
[EXTERNAL_DOWNLOADS]: The skill interacts with the GitHub API (gh-api) to retrieve PR and commit metadata for squash-merge detection. This network activity targets a well-known, trusted service and is limited to metadata retrieval without executing remote content.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of Git conflict markers and conflicted file content. This is the primary function of the tool, and the risk is mitigated by the use of structural merge tools and manual review steps defined in the protocol.
[FILE_SYSTEM_WRITE]: The skill provides instructions to manage the ~/.gitattributes file to register the mergiraf merge driver. This is a standard configuration task for the associated tooling and does not constitute a persistence mechanism.

melt