The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill explicitly directs the agent to "re-run it with escalated permissions" if a runtime_access_denied error is encountered. This instruction encourages the agent to bypass sandbox restrictions and gain higher system privileges.
[INDIRECT_PROMPT_INJECTION]: The skill utilizes a dynamic loading mechanism where the primary usage instructions and coordinator logic are fetched at runtime via the ORCA skills get orchestration command. This creates a vulnerability where the tool's output could potentially influence the agent's behavior through embedded instructions.
Ingestion points: The output of the shell command ORCA skills get orchestration as described in SKILL.md.
Boundary markers: Absent. The instructions do not specify any delimiters or warnings to treat the fetched content as untrusted data.
Capability inventory: The agent is granted capabilities to execute shell commands, manage worktrees, control terminals, and open the built-in browser via the orca CLI.
Sanitization: Absent. There are no instructions provided to validate, filter, or escape the content returned by the binary before the agent acts upon it.
[COMMAND_EXECUTION]: The skill centers on resolving and executing various local binaries (orca, orca-ide, orca-dev) and managing system state via shell commands. It includes logic to select the executable based on environment variables like ORCA_CLI_COMMAND or ORCA_DEV_REPO_ROOT.

orchestration