The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to interact with git repositories, including git diff for analysis and git push to synchronize the Obsidian vault with its remote origin.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from git diff outputs, creating a surface for indirect prompt injection if the code being reviewed contains malicious instructions.
Ingestion points: The agent reads arbitrary code changes using git diff main...HEAD in the project directory.
Boundary markers: There are no explicit instructions or delimiters used to separate the code diff from the agent's instructions, nor are there commands to ignore embedded instructions.
Capability inventory: The skill possesses the capability to execute shell commands and perform file system writes and edits in the Obsidian vault.
Sanitization: No sanitization or filtering is performed on the diff content before it is processed by the agent.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill writes review findings to a local filesystem and subsequently pushes those changes to a remote Git repository. While this is the intended synchronization mechanism for Obsidian users, it involves transmitting data derived from private codebases to a remote server (the vault's Git origin).

ai-code-review