The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill explicitly instructs the agent to "re-run it with escalated permissions" if a runtime_access_denied error occurs due to sandbox restrictions. This is a direct instruction to bypass security boundaries.
[COMMAND_EXECUTION]: The skill is designed to resolve and execute local binaries (orca, orca-ide, or orca-dev) based on environment variables or specific system contexts.
[DYNAMIC_EXECUTION]: The instruction ORCA skills get orca-cli is used to load a "version-matched guide" at runtime. This dynamically alters the agent's instructions based on the output of a CLI tool, which is a form of runtime code/instruction loading.
[INDIRECT_PROMPT_INJECTION]: The skill lacks boundaries for the content it ingests from the orca CLI tool, creating a vulnerability surface where the tool's output could influence agent behavior.
Ingestion points: The output of the command ORCA skills get orca-cli (in SKILL.md).
Boundary markers: Absent. The agent is told to "load the version-matched guide" and then "Use the selected executable's --help for commands or flags the guide does not cover."
Capability inventory: The skill can execute arbitrary sub-processes via the resolved orca executable and is instructed to seek escalated (root/admin) permissions if blocked.
Sanitization: Absent. The agent treats the output of the CLI command as a trusted guide without validation or escaping.

orca-cli