The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed around executing the orca CLI (including variants like orca-dev and orca-ide) to manage worktrees, terminals, repositories, and other system resources.
[PRIVILEGE_ESCALATION]: The instructions explicitly direct the agent to bypass sandbox restrictions. If a command fails with a runtime_access_denied error, the skill instructs the agent to 're-run it with escalated permissions'. This encourages the agent to use tools like sudo to bypass environmental security boundaries.
[INDIRECT_PROMPT_INJECTION]: The skill relies on dynamic context by instructing the agent to run ORCA skills get orca-cli and treat the output as an authoritative guide for subsequent actions.
Ingestion points: The output of the shell command ORCA skills get orca-cli is ingested into the agent's context (referenced in SKILL.md).
Boundary markers: None. The dynamic output is not delimited, nor is there a warning to ignore instructions embedded within the guide.
Capability inventory: The skill provides capabilities to manipulate worktrees, filesystems, terminals, and an embedded browser.
Sanitization: There is no evidence of sanitization or validation performed on the CLI output before the agent processes it as instructions.

orca-cli