The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch and install the OpenShell dependency from the NVIDIA/OpenShell repository using the npx skills add command if it is not already present.
[COMMAND_EXECUTION]: The skill contains a shell command npx skills add NVIDIA/OpenShell intended for environment setup and dependency management when an isolated worker is required.
[INDIRECT_PROMPT_INJECTION]: The skill defines logic for processing external AI tasks and routing them to specific execution boundaries. While it acts as a security filter, the task inputs themselves represent an attack surface for indirect prompt injection.
Ingestion points: External AI task descriptions are processed by the routing logic defined in SKILL.md.
Boundary markers: The skill explicitly defines boundaries (Route A, B, and C) and recommends using isolated workspaces and brokered credentials to mitigate risks.
Capability inventory: The skill has the capability to execute shell commands (npx skills add) for setup purposes.
Sanitization: The skill advocates for isolation techniques, such as using disposable workers and explicit network/API scoping, to sanitize the execution environment.

safe-run