The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill contains instructions that direct the agent to bypass environment security controls by escalating permissions.
Evidence: In SKILL.md, the text states: "If it fails with runtime_access_denied, your sandbox blocked the connection: re-run it with escalated permissions".
[COMMAND_EXECUTION]: The skill performs runtime resolution of executable paths based on environment variables and context, leading to the execution of various binaries.
Evidence: Instructions in SKILL.md guide the agent to select between orca, orca-dev, or orca-ide based on the presence of ORCA_CLI_COMMAND or ORCA_DEV_REPO_ROOT environment variables.
[INDIRECT_PROMPT_INJECTION]: The skill relies on fetching operational instructions at runtime from a tool's output, which the agent is then instructed to follow.
Ingestion points: The output of the command ORCA skills get orchestration (and its variants) is used as a reference for agent behavior.
Boundary markers: Absent. The skill instructs the agent to "read the named bundled reference before acting" without specifying delimiters or safety warnings for the fetched content.
Capability inventory: The agent is granted capabilities to execute shell commands, manage worktrees, and control a browser via the Orca CLI binaries.
Sanitization: None. The skill assumes the fetched guide is authoritative and does not suggest validation of the content before execution.
[DYNAMIC_EXECUTION]: The skill dynamically assembles its execution environment by resolving command paths at runtime and loading instruction sets from external tool outputs.

orchestration