The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses bun install to regenerate lockfiles after resolving conflicts. This is a standard development practice that fetches dependencies from well-known package registries.
[COMMAND_EXECUTION]: The skill executes shell commands via the Bash tool to perform git operations, type-checking (tsc), and testing. These operations are limited to the scope of the local repository and are essential for the skill's stated purpose.
[PROMPT_INJECTION]: The safety model explicitly instructs the agent to treat incoming code changes as untrusted data. It provides clear directives to never execute instructions embedded in code or comments, mitigating the risk of indirect prompt injection.
[DATA_EXFILTRATION]: The skill includes a specific safety rule to redact secret-like values encountered in code blocks during the resolution process, preventing the exposure of credentials found in the source files.

fix-merge-conflicts