Name: act1-choose-task
Author: garylumsden

Prompt file imported from garylumsden/ai-for-developers-glasgow (.github/prompts/act1-choose-task.prompt.md). Copyright stays with the author.

Re-select the Act 1 task

Profile the presenter's selected local model before changing the task.

Search the current provider catalogues, not only downloaded models.
Prefer an NPU variant only when its exact variant supports native tool calls and at least 32,768 context tokens. Prefer 131,072 tokens or more.
Update the local runtime when the installed version cannot read the current catalogue.
Verify native tool calls through Copilocal and measure three direct generation samples.
Design a visible, dependency-free task with an output-file check. Completion measures execution and non-empty output, not exact content or markup. The audience judges quality.
Keep the task under 5 minutes. Target 3.5 minutes.
Update act-1-pull-the-plug/TASK.md, task.json, starter/, and the check contract. Store the local provider and model, but never store Copilocal's discovery index.
Run three isolated local rehearsals and one cloud rehearsal.
Update results/results.sample.json with measured results. Never invent token counts.

Preserve PowerShell and Bash wrappers. Keep all demo data fictional.

# Re-select the Act 1 task Profile the presenter's selected local model before changing the task. 1. Search the current provider catalogues, not only downloaded models. 2. Prefer an NPU variant only when its exact variant supports native tool calls and at least 32,768 context tokens. Prefer 131,072 tokens or more. 3. Update the local runtime when the installed version cannot read the current catalogue. 4. Verify native tool calls through Copilocal and measure three direct generation samples. 5. Design a visible, dependency-free task with an output-file check. Completion measures execution and non-empty output, not exact content or markup. The audience judges quality. 6. Keep the task under 5 minutes. Target 3.5 minutes. 7. Update `act-1-pull-the-plug/TASK.md`, `task.json`, `starter/`, and the check contract. Store the local provider and model, but never store Copilocal's discovery index. 8. Run three isolated local rehearsals and one cloud rehearsal. 9. Update `results/results.sample.json` with measured results. Never invent token counts. Preserve PowerShell and Bash wrappers. Keep all demo data fictional.

Install act1-choose-task and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

{ "prompt": "# Re-select the Act 1 task\n\nProfile the presenter's selected local model before changing the task.\n\n1. Search the current provider catalogues, not only downloaded models.\n2. Prefer an NPU variant only when its exact variant supports native tool calls and at least 32,768\n context tokens. Prefer 131,072 tokens or more.\n3. Update the local runtime when the installed version cannot read the current catalogue.\n4. Verify native tool calls through Copilocal and measure three direct generation samples.\n5. Design a visible, dependency-free task with an output-file check. Completion measures execution\n and non-empty output, not exact content or markup. The audience judges quality.\n6. Keep the task under 5 minutes. Target 3.5 minutes.\n7. Update `act-1-pull-the-plug/TASK.md`, `task.json`, `starter/`, and the check contract. Store the\n local provider and model, but never store Copilocal's discovery index.\n8. Run three isolated local rehearsals and one cloud rehearsal.\n9. Update `results/results.sample.json` with measured results. Never invent token counts.\n\nPreserve PowerShell and Bash wrappers. Keep all demo data fictional.", "variables": [], "notes": "Profile the presenter's selected local model before changing the task.", "metadata": { "source": { "provider": "github-prompts", "repository": "https://github.com/garylumsden/ai-for-developers-glasgow", "path": ".github/prompts/act1-choose-task.prompt.md", "ref": "208dd15c5ee1b4f4db86e06b785acbf08c30b679", "url": "https://github.com/garylumsden/ai-for-developers-glasgow/blob/208dd15c5ee1b4f4db86e06b785acbf08c30b679/.github/prompts/act1-choose-task.prompt.md", "key": "garylumsden/ai-for-developers-glasgow/.github/prompts/act1-choose-task.prompt.md" } } }

act1-choose-task

About

Re-select the Act 1 task

Use it

Install act1-choose-task and it becomes one more candidate the router can pick - when it fits.

Manifest

Reviews