The Agent Skills Directory

[COMMAND_EXECUTION]: The deployment script scripts/deploy.py and the deployed project hooks (guard-outline-before-prose.sh, validate-story-commit.sh, etc.) execute local shell commands including git, python, and node. These commands are used to manage project directories, verify installation integrity, and enforce workflow requirements such as checking for the existence of chapter outlines before allowing writing to proceed.
[EXTERNAL_DOWNLOADS]: The session-start.sh hook performs an update check by using curl to retrieve the latest release information from the author's public repository on GitHub (api.github.com/repos/Chained1001/mo-shu/releases/latest). This is a legitimate functional feature that uses a trusted service to fetch non-executable version metadata.
[INDIRECT_PROMPT_INJECTION]: The moshu-researcher agent is instructed to use the agent-browser tool to retrieve external data for novel research, which presents a surface for potential indirect prompt injection from third-party web content. This risk is mitigated within the agent's instructions by explicit requirements for objective analysis, source cross-verification, and instructions to disregard embedded commands in retrieved material.
[PERSISTENCE]: The skill maintains its functional presence by modifying the Claude Code configuration file .claude/settings.local.json to register its shell scripts as automated lifecycle hooks. This method of persistence is a standard feature of the target platform and is required for the toolset's automated validation features to operate across sessions.

moshu-setup