agentleFS
Cody-SimsCody-Sims/agent-skills/skills/copilot-cloud-agent/SKILL.md

copilot-cloud-agent

Skill0 starsChanged 4 months agoMITView on GitHub

Launches and monitors GitHub-hosted Copilot tasks with explicit models and durable task/PR IDs. Use when starting cloud agents, assigning issues to hosted Copilot, selecting GPT-5.6 Sol, or checking cloud-task status; not for local CLI subagents or Git worktree fleets.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use genuine GitHub-hosted tasks, never relabeled local workers.

    line 13
  2. Distinctive:

    If a hard execution-time or spend ceiling is required, verify enforceable hosted support before launch.

    line 21
  3. Very specific:

    Status only: read known IDs directly for state/model/refs/artifacts.

    line 23
  4. Distinctive:

    Authorized launch/assignment: read applicable hosted controls.

    line 25
  5. Very specific:

    Prefer authenticated gh api.

    line 33
  6. Distinctive:

    Verify PR existence before reporting its URL.

    line 40

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Copilot Cloud Agent

Use genuine GitHub-hosted tasks, never relabeled local workers. Loading/status requests authorize no launch, assignment, resumption, cancellation, or merge.

Workflow

  1. Record repository/action, known IDs, base/head refs, bounded prompt, PR choice, and requested model. Confirm hosted scope and plan/organization policy. Omitted model means Auto, not a specific GPT choice. If a hard execution-time or spend ceiling is required, verify enforceable hosted support before launch; otherwise report blocked and seek a decision.
  2. Status only: read known IDs directly for state/model/refs/artifacts. Discover bounded matches only if the ID is missing; never relaunch for status.
  3. Authorized launch/assignment: read applicable hosted controls. Verify current API shape/requested model from linked GitHub sources before each launch; reuse same-operation evidence, not stale snapshots. Check matching tasks/PRs by repository/refs/objective; bound pagination. Reuse or wait for matching work; report/block launch if discovery cannot exclude duplicates.
  4. Tasks use POST /agents/repos/{owner}/{repo}/tasks: prompt and applicable documented base_ref, head_ref, model, custom_agent, create_pull_request. Prefer authenticated gh api; never expose credentials. Issue workflows use the documented assignment API/model field, not a model-dropping helper.
  5. Send the requested model explicitly, e.g. model: "gpt-5.6-sol" when available. Surface validation/policy failures; never silently fall back. After uncertain POST outcomes inspect matching tasks before retry; avoid duplicate creation.
  6. Record task ID/repository URL immediately and read back state/model/refs/artifacts. Preserve exact provider-prefixed model values; report missing fields/mismatches. Verify PR existence before reporting its URL.
  7. Bound known-ID polling: default five reads or five minutes, whichever first; back off 15/30/60 seconds, honor Retry-After, and stop at the budget. Read concise deltas; bounded logs only for triage. Stop for completed, failed, timed_out, cancelled, or needed user input. queued, in_progress, idle, and waiting_for_user are not completion. Budget exhaustion means active/blocked; no unauthorized resume/relaunch/retry. A polling budget does not cap hosted compute/spend or authorize cancellation.
  8. Review PR/code and repository-required checks before an authorized merge. Completed sessions prove neither correctness nor approval.

Hosted versus local controls

  • The API snapshot documents model; it does not currently document effort, reasoning_effort, context, context_window, or token-window fields. Recheck current sources when answering control questions; never invent fields/CLI flags.
  • Local CLI controls do not configure hosted tasks. Sol selection guarantees neither effort, long-context mode, nor exact 1M allocation.

Output

Return host/repository, task ID/URL/state, requested/observed model, base/head refs, verified PR URL, budget/results, restrictions, and pending validation. No transcript or credentials; distinguish queued, active, blocked, failed, and completed.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.