The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies extensively on executing shell commands to manage the project lifecycle. This includes using corepack pnpm for package management, node for running audit scripts and version checks, and makecli for deployment and artifact diffing.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to perform dependency installations using corepack pnpm install --frozen-lockfile. While this fetches external code from registries, it follows security best practices for reproducible builds.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface as it ingests and processes untrusted project data to drive its logic.
Ingestion points: apps/package.json, apps/pnpm-workspace.yaml, apps/ui/package.json, apps/service/package.json, and .nvmrc (all located in SKILL.md).
Boundary markers: Absent; there are no specific delimiters or warnings to the agent to ignore instructions embedded within these manifests.
Capability inventory: The agent can execute subprocesses via corepack pnpm, node, and makecli, and has read/write access to the filesystem to manage build artifacts in apps/ui/dist and apps/service/dist.
Sanitization: Absent; the skill does not specify validation or sanitization of the values read from the JSON/YAML manifests before using them in shell commands.
[DYNAMIC_EXECUTION]: The skill uses node -e to perform a runtime version check. This is a low-risk, standard practice for ensuring environment compatibility before starting build processes.

make-app-runtime