The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill is designed to process untrusted data from source code changes and pull request descriptions to automate documentation tasks.
Ingestion points: Data enters the agent context via 'git diff' output and by reading source files in the 'packages/next/src/' directory.
Boundary markers: Absent. There are no delimiters or specific instructions provided to the agent to ignore or isolate instructions embedded within the code or markdown files being processed.
Capability inventory: The skill has the authority to execute shell commands ('git', 'pnpm') and perform file-system write operations to update MDX documentation.
Sanitization: None. The skill lacks validation or sanitization of the content extracted from the code changes before using it to generate documentation.
[Remote Code Execution] (HIGH): The workflow includes the execution of 'pnpm lint' and 'pnpm prettier-fix'.
Risk: In the context of reviewing PRs from external contributors, a malicious actor could modify the 'package.json' file to include a malicious command within the 'lint' or 'prettier-fix' scripts. When the agent executes these commands as part of its validation step, it would trigger arbitrary code execution in the host environment.

update-docs