The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The instructions contain a command to install the plannotator CLI by piping a remote script directly to the shell: curl -fsSL https://plannotator.ai/install.sh | bash. This represents a high-privilege execution pattern, although it originates from the vendor's primary domain.
[DATA_EXFILTRATION]: The plannotator guide share command transmits the changeset diff and the generated guide to an external hosting service (guides.show). While the documentation notes that this is end-to-end encrypted, it involves moving local repository data to a third-party server.
[COMMAND_EXECUTION]: The skill requires several local shell operations to function, including generating patches via git diff, interacting with GitHub via gh pr view, and invoking the plannotator CLI for export and sharing tasks.
[EXTERNAL_DOWNLOADS]: The skill initiates network requests to plannotator.ai to retrieve the installation script and to guides.show when utilizing the sharing functionality.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository's diff (guide.patch) and uses it to generate descriptive chapters.
Ingestion points: Reads local diff files and PR descriptions.
Boundary markers: None explicitly defined in the prompt instructions to separate untrusted diff content from the agent's logic.
Capability inventory: Subprocess execution for git, gh, and the vendor CLI, as well as network access for installation and sharing.
Sanitization: No specific sanitization or escaping of the diff content is described before the agent processes it to write the guide.

guided-review