The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses shell commands through npx --package @notis_ai/cli@latest to manage the application lifecycle, including initializing projects, starting development servers, building bundles, and performing deployments.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download the Notis CLI tool dynamically via npx and to install project dependencies using npm install. It also retrieves application scaffolds from a public registry to use as starting points for new projects.
[DYNAMIC_EXECUTION]: The architecture relies on building Vite and React projects into ES module bundles which are then dynamically loaded and rendered as React components within the Notis portal tree.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, creating a potential attack surface where malicious input could influence agent actions.
Ingestion points: The skill pulls source code from a public registry via notis apps init --from <slug> and reads information from notis apps scaffolds list. It also processes data retrieved from databases through tools like LOCAL_NOTIS_DATABASE_QUERY.
Boundary markers: The skill requires explicit user approval of required_capabilities before installation and uses specific tokens (e.g., workspace_databases_read) to scope access.
Capability inventory: The environment allows for shell command execution (npx), filesystem operations (writing app code), and network operations (CLI communication with Notis servers).
Sanitization: While the skill emphasizes user approval for capability grants, it does not explicitly define sanitization or escaping procedures for data ingested from external registries or databases.
[DATA_EXPOSURE]: The skill interacts with sensitive local files such as ~/.notis/state.json, which contains development identities and environment links, to manage the application state across sessions.

notis-apps