delegation

While delegation mode is on, you orchestrate — think, plan, brief, verify — and subagents execute, on the seat model-selector chooses, unless you run them on the frozen main seat with no override. This skill turns the mode on or off. Your own model and reasoning effort stay the user's move — whatever your harness offers for changing them is theirs to run, never yours.

Run uv run "$HERE/scripts/invoke.py" — $HERE is the directory that holds this SKILL.md — with everything the user typed after /delegation, verbatim and however many lines, on stdin. Exit 0: do what it prints. On any other exit, if you introduced a known construction error and can correct it while preserving the user's request and authority, account for effects already produced, submit the corrected invocation through the same shim, and continue from the failed boundary; a refusal before the operation starts consumes no operation. Otherwise show what it printed to the user verbatim and stop. Never repair input the user supplied, or automatically retry exact help, an unmet dependency, an unrelated failure, or a failure whose origin or valid correction is unknown.

Arguments

• --project and --user name the two persistent scopes. Without either, the scope is this session, which has no name of its own.
• --yes — write the persistent scope without waiting for a yes. status asks nothing.
• The session is the only togglable scope, and the bare invocation — an empty path — flips its current verdict.
• status with no scope flag reports all three scopes.

The mode

$HERE/references/mode.md is the single source of truth for what the mode says. Read and adopt it as it stands, with <model-selector> resolved as step 2 says. Persistent scopes follow persist.md: they copy the mode and fence to two companion files — verbatim, the placeholder included, so a committed trio carries no one machine's path — and keep only an @ pointer block in the context file.

Two states, no third: