The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing various shell commands including git rebase, git ls-remote, and git fetch. These are core to the skill's functionality and are authorized by the user's request to rebase a branch.
[DATA_EXFILTRATION]: The skill performs network operations using git ls-remote --symref and git fetch to interact with remote repositories. The target domains are not fixed and are determined by the local Git configuration of the environment where the skill is executed.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Git repository, particularly during the conflict resolution phase. This creates a surface where an attacker could embed malicious instructions in commit messages or file content to influence agent behavior.
Ingestion points: Workflow step 5 in SKILL.md involves inspecting conflict markers and diff stages (:1:, :2:, :3:) from files within the repository.
Boundary markers: The skill does not implement specific delimiters to isolate repository content from instructions, relying instead on standard Git conflict markers.
Capability inventory: The skill can execute shell commands (git), modify local files (during conflict resolution), and initiate network connections (git fetch).
Sanitization: There is no evidence of sanitization or filtering of external content before it is processed by the agent's logic.

rebase-assistant