The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill contains an explicit instruction for the agent to seek higher privileges: 'If it fails with runtime_access_denied, your sandbox blocked the connection: re-run it with escalated permissions'. This encourages the agent to use commands like sudo to bypass local environment restrictions.
[COMMAND_EXECUTION]: The skill relies on the execution of local binaries (orca, orca-dev, or orca-ide) determined by environment variables like ORCA_CLI_COMMAND and ORCA_DEV_REPO_ROOT. It warns about name collisions with system utilities (e.g., the GNOME Orca screen reader).
[INDIRECT_PROMPT_INJECTION]: The skill uses a 'discovery stub' pattern where it runs ORCA skills get orca-cli to fetch instructions from the output of a local command. This introduces an attack surface where the tool's output dynamically provides the primary instructions for the agent's behavior.
Ingestion points: Instructions are loaded via the output of the command ORCA skills get orca-cli in SKILL.md.
Boundary markers: None provided; the agent is instructed to load and follow the version-matched guide directly.
Capability inventory: The skill facilitates worktree management, terminal interaction, and browser control via the orca CLI.
Sanitization: There is no apparent sanitization or validation of the guide content retrieved from the executable.

orca-cli