This document describes Rally's wrapper entrypoint for harness slash commands.
Install wrapper artifacts for one harness:
rally skill install --target codexInstall for all supported harnesses:
rally skill install --target allSupported targets:
codex->~/.codex/skills/rally/SKILL.mddroid->~/.factory/skills/rally/SKILL.mdpi->~/.pi/agent/skills/rally/SKILL.mdclaude-code->~/.claude/skills/rally/SKILL.md
Each wrapper is Rally-managed content and delegates back to:
rally skill agent-howto "$@"The generated /rally wrapper is intentionally thin:
- accepts wrapper arguments unchanged
- forwards arguments to
rally skill agent-howto - keeps the mutable prompt in Rally source code (
prompts/skill-agent-howto.md) - keeps orchestration logic centralized in Rally
Examples:
/rally/rally build/rally build reviewer/rally <invite-code>
agent-howto prints the current instruction loop and exact follow-up run command:
rally skill agent-howto
rally skill agent-howto implement todos/create-todo.md implementrun and exec are equivalent:
rally skill run
rally skill execResolution precedence:
- explicit args (
--session,--as, positional selectors) - invite token (for example
rly:<session>:<agent>,rally://<session>/<agent>) - saved workspace context (
~/.rally/skill-context.json) - interactive fallback prompt
Backward compatibility note:
- If
skill-context.jsonis missing, Rally falls back to reading legacycommand-context.json.
Inspect current installation health:
rally skill doctor --target allDoctor reports per target:
- environment/path availability
- wrapper presence
- managed/unmanaged file state
- managed content drift
For preview mode:
rally skill doctor --target all --dry-runRemove Rally-managed wrappers only:
rally skill uninstall --target allUnmanaged user files are never removed.
Preview removal decisions:
rally skill uninstall --target all --dry-runWrappers map to Rally core loop behavior through rally skill run, which then joins/polls existing session workflows:
- registration path:
rally join --session <name> --as <agent>(as needed) - instruction polling path:
rally next --session <name> --as <agent> - review flow remains canonical under
rally build checkpoint|review - plan actions remain canonical under
rally plan ...