The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructions direct the agent to run eval "$(node ...)". The underlying script scripts/npm-token.mjs generates shell export commands using strings (authToken and registry) retrieved from a remote server (TARGET_XGENT_PLATFORM). This creates a potential command injection surface if the remote server provides a specially crafted response designed to escape the string literals in the generated shell code.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, presenting an attack surface for indirect prompt injection. Ingestion points: The skill reads app.manifest.json in scripts/preflight.mjs and parses API responses from the XGENT portal in scripts/npm-token.mjs. Boundary markers: No specific delimiters or instructions to ignore embedded instructions are used when handling the manifest or API data. Capability inventory: The skill can execute shell commands via npx, node, and tar, and perform network operations via fetch. Sanitization: The references/embed-csp.mjs script performs strict sanitization of CSP fields using regex. scripts/preflight.mjs validates version and key formats. However, scripts/npm-token.mjs does not sanitize server-provided strings used in shell command generation.
[COMMAND_EXECUTION]: The skill executes various system commands, including tar for packaging (via spawnSync in scripts/preflight.mjs) and npx for running the @xgent/release-cli tool. These operations are essential for the skill's build and release tasks.
[EXTERNAL_DOWNLOADS]: The skill fetches the @xgent/release-cli tool and other SDK dependencies (e.g., @xgent/shared, @xgent/portal-sdk) from a private npm registry. These are vendor-owned resources from the xgent-ai organization.

xgent-app-release