The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the stagereview package globally via npm install -g or execute it directly using npx. This package is hosted on the public NPM registry and its source code is maintained in an external repository (github.com/ReviewStage/stage-cli).
[COMMAND_EXECUTION]: The skill relies on the execution of the stage-chapters command, which performs local system operations including analyzing the git repository, detecting local changes, and initiating a local web server (typically on port 3000) to host the review interface.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because it ingests untrusted data in the form of local source code and git diffs which are then processed by an AI agent.
Ingestion points: The tool reads staged, unstaged, and untracked file contents and git diffs from the local repository as specified in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or specific warnings to prevent the agent from following instructions embedded within the code being reviewed.
Capability inventory: The skill executes shell commands (stage-chapters, git) and interacts with the local network via a web server.
Sanitization: There is no mention of sanitization or filtering of the code content before it is presented to the agent for analysis.
[PERSISTENCE]: The documentation suggests integrating the tool into a git pre-commit hook. While a standard development practice, this configures the system to automatically execute the stage-chapters script whenever a user attempts to commit changes in the repository.

stage-cli-code-review