The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation requires the agent to execute shell commands (orx skills add, orx templates add) to modify its own capability set and stored templates.
[INDIRECT_PROMPT_INJECTION]: The skill enables the ingestion of local files that can contain instructions, which could influence the agent's behavior if malicious content is included in the added files.
Ingestion points: File path arguments passed to the orx commands as described in SKILL.md (e.g., ./SKILL.md, ./research-workflow.zip).
Boundary markers: The instructions do not define boundary markers or instruct the agent to ignore instructions embedded within the ingested files.
Capability inventory: Execution of the orx suite of tools to register persistent skills and templates.
Sanitization: No validation or sanitization of file contents is mentioned or required in the skill instructions.

orx-customize