The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the !command syntax to automatically execute shell commands when the skill is loaded into the agent context. These commands include cat, find, head, and xargs to gather project information such as package manifests, color configurations, icons, and metadata. While these are legitimate discovery commands for a development tool, they execute silently without explicit user approval for each call during the skill activation phase.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from the user's local project files to infer app categories, features, and brand details.
Ingestion points: Multiple files including README.md, package.json, app.json, pubspec.yaml, and Fastlane metadata directories.
Boundary markers: Absent. The instructions do not specify delimiters or instructions to ignore embedded commands within the ingested text.
Capability inventory: The skill is capable of performing file system writes (saving .appshot-context.json) and executing shell commands for project building and development servers.
Sanitization: No sanitization or validation of the content of these external files is performed before the agent processes them for high-level decision making (like category inference).
[COMMAND_EXECUTION]: The skill documentation provides shell command patterns for the agent to execute in order to scaffold the appshot-video/ directory and manage the development environment (e.g., npm run dev, npm run build).

appshot-core