The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and acting upon untrusted data from external sources.
Ingestion points: The agent reads external content via gh run view --log-failed (CI logs) and gh api (automated bot reviews from cubic-dev-ai[bot]).
Boundary markers: No explicit delimiters or instructions are used to separate user-provided task instructions from the potentially adversarial content found in logs or review bodies.
Capability inventory: The agent possesses high-privilege capabilities including the ability to execute shell commands (git, gh, bun), write files in the worktree, and push code to the remote repository.
Sanitization: The skill lacks sanitization or validation logic to ensure that processed logs or reviews do not contain malicious instructions intended to manipulate the agent's behavior.
[COMMAND_EXECUTION]: The skill performs automated execution of commands that interact with the local filesystem and remote repository.
It executes bun install, bun test, and bun run build within the newly created worktree. This executes code defined within the repository's configuration files, which is a necessary but high-capability operation for a development agent.

work-with-pr