Menu
Plan the next action for the top-scored grant opportunity using a Gemma 4 function call.
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
Use when working in Coop and needing the local app, receiver PWA, docs, API/signaling server, or extension watcher.
Browser and visual verification for Coop app, extension, popup, sidepanel, and receiver flows. Use when UI changes need rendered proof, console inspection, screenshots, Playwright checks, or local browser debugging.
Debugging & Troubleshooting - systematic root cause investigation with hypothesis testing and evidence collection. Use when the user reports a bug, encounters an error, sees unexpected behavior, or says 'debug this' or 'investigate this issue'.
Quality gate and regression monitor. One-shot (`/monitor`) or continuous (`/loop 5m /monitor`). Scope with --tests, --build, or run all checks by default.
Performance profiling and optimization - bundle analysis, Lighthouse CI, React Profiler, memory leaks, service worker cache efficiency. Use for performance investigations, optimization, and budget enforcement.
Planning & Execution - create structured implementation plans, check progress, execute in batches, manage lifecycle. Use when the user says 'plan this', asks to break down a feature into steps, or needs a phased implementation strategy before coding.
| name | grant-action-planner |
| description | Plan the next action for the top-scored grant opportunity using a Gemma 4 function call. |
Pick the strongest scored opportunity in the latest fit-scorer pass, then call exactly one of:
draft_application_outline — propose section headers for an application response.add_to_coop_calendar — surface the deadline so the coop sees it on the next review.request_member_input — name members whose context is needed before drafting.Use the coop purpose, recent decisions, and any attached audio meeting context as fit signal. Keep
the rationale short and tied to the opportunity. The model is expected to emit exactly one
<tool_call> payload — the runner discards trailing prose.