Menu
Use `al dispatch` only when the user names an external dispatch target or another skill explicitly requires dispatch. Do not use it for generic subagent, second-agent, or fresh-context requests; use the built-in subagent instead.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Explicit-only. Run a named autonomous mode, repeatedly selecting, implementing, and shipping coherent work while preserving anything blocked.
Use globally installed `playwright-cli` for browser automation, UI inspection, screenshots, Playwright test debugging, or test generation. Trigger when a task needs browser interaction or test repair. Do not use for generic tests, web search, non-browser docs, or API-only work.
Explicit-only. Align a feature specification, produce a reviewed plan, complete the local work, and ship the pull request.
Explicit-only. Review a plan/task/context artifact set and report material findings.
Explicit-only. Ship one pull request, observe hosted CI and review feedback, follow PR policy, reply to every eligible comment, and request merge authorization before merging and cleaning up.
Explicit-only. Verify completed work against its authoritative contract and report coverage, working-code evidence, material gaps, and the completion verdict.
| name | agent-dispatch |
| description | Use Agent Dispatch MCP tools only when the user names an external dispatch target or another skill explicitly requires dispatch. Do not use it for generic subagent, second-agent, or fresh-context requests; use the built-in subagent instead. |
| compatibility | Requires the built-in `agent-layer` MCP server and a configured provider. |
| allowed-tools | mcp__agent-layer__dispatch_options mcp__agent-layer__dispatch_start mcp__agent-layer__dispatch_wait mcp__agent-layer__dispatch_continue mcp__agent-layer__dispatch_cancel Bash(cat:*) |
If the MCP tools are unavailable, report the missing server; do not substitute command-line calls.
dispatch_options; resolve the provider and overrides. Ask if
ambiguous.dispatch_start once with exactly one prompt source and retain its
handle. Do not replace active work.dispatch_wait with the handle. On running or an interrupted wait,
wait again with the same handle; provider work remains active. On
completed, read the Markdown file at result_path; report failed or
cancelled.For parallel work, call dispatch_start once per independent conversation and
retain each handle. Call dispatch_wait for those handles in parallel when supported.
After a terminal result, use dispatch_continue only for useful follow-up,
requested information, or corrective action within the current scope. It
preserves the provider conversation context. Pass the same handle and exactly
one prompt source, then call dispatch_wait again.
Use dispatch_start when fresh context is required, not dispatch_continue.
dispatch_cancel permanently stops active provider work. Call it only when the
user explicitly requests cancellation or an active skill explicitly instructs
you to abandon the dispatch.