The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes gt commands to perform complex version control operations, including branch creation, rebasing, and syncing stacks.
[EXTERNAL_DOWNLOADS]: The gt upgrade command allows the agent to fetch and install updates for the Graphite CLI directly from official remote servers.
[REMOTE_CODE_EXECUTION]: The gt upgrade functionality facilitates the replacement of the local CLI binary with a remote version, a high-privilege action performed from a well-known service.
[CREDENTIALS_UNSAFE]: The gt auth command is used for configuring GitHub authentication tokens, necessitating secure management of sensitive credentials by the agent to prevent exposure.
[PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection from external data sources. -- Ingestion points: The skill reads Pull Request bodies via gt info --body and commit messages through gt log, which are then processed within the agent's prompt context. -- Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from following instructions potentially embedded in these external data fields. -- Capability inventory: The skill possesses significant capabilities, such as modifying repository contents (gt modify) and creating/submitting Pull Requests (gt submit), which could be leveraged if an injection occurs. -- Sanitization: No filtering or sanitization of the ingested text is performed before it reaches the agent.

graphite