The Agent Skills Directory

[DYNAMIC_EXECUTION]: The analysis tools provided in the skill utilize dynamic code loading and execution to audit the local project environment.
The script scripts/lib/app-modules.mjs uses module.registerHooks and dynamic import() to load the project's TypeScript modules for inspection.
The utility scripts/lib/link-facts.mjs uses spawnSync to execute a Node.js child process that dynamically evaluates the project's app.config.ts to extract variant identity and link configuration.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process untrusted deep link data from the operating system, which is a potential surface for injection attacks.
Ingestion points: External URLs enter the agent context through src/app/+native-intent.tsx.
Boundary markers: The system implements a strict allow-list policy (LINK_ROUTES) that defines permitted paths and parameters.
Capability inventory: The logic is contained to route resolution and does not permit arbitrary command execution or unauthorized file system access based on link inputs.
Sanitization: The gate logic in route-incoming-link.ts performs re-encoding of URI components and validates path segments against regex-based kind definitions to prevent malformed or malicious payloads.

deep-links-and-universal-links