Install any skill in seconds. Free to start, no credit card required.
Get Started Free →GitHub Copilot CLI as optional zero-cost provider via copilot -p programmatic mode
.claude/skills/hashgraph-online-skill-copilot-provider/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -22% | 0% |
> Host: Codex CLI — This skill was designed for Claude Code and adapted for Codex. > Cross-reference commands use installed skill names in Codex rather than /octo:* slash commands. > Use the active Codex shell and subagent tools. Do not claim a provider, model, or host subagent is available until the current session exposes it. > For host tool equivalents, see skills/blocks/codex-host-adapter.md.
GitHub Copilot CLI (GA since Feb 2026) serves as an optional provider in the Claude Octopus multi-LLM ecosystem. Integration uses the official copilot -p programmatic mode, not reverse-engineered API endpoints.
Core principle: Copilot supplements existing providers for research and exploration at zero additional cost (uses existing GitHub Copilot subscription). Each prompt counts as one premium request against your subscription quota.
Agent types: copilot (general), copilot-research (research-focused)
bash# Check copilot CLI is available if ! command -v copilot &>/dev/null; then # Copilot CLI not installed — silently skip return 0 fi
Graceful degradation: When Copilot CLI is unavailable or unauthenticated, silently skip. Other providers continue to operate normally.
Copilot CLI checks credentials in this precedence order:
COPILOT_GITHUB_TOKEN env var (highest priority — fine-grained PAT with "Copilot Requests" permission)GH_TOKEN env varGITHUB_TOKEN env varcopilot login)gh) authentication fallbackOption 1: Interactive login (recommended for local dev)
bashcopilot login
Option 2: Fine-grained PAT (recommended for CI/automation)
bashexport COPILOT_GITHUB_TOKEN="github_pat_..."
Option 3: Reuse existing gh auth If gh auth login is already configured, Copilot CLI will use it automatically.
Note: Classic PATs (ghp_*) are NOT supported. Use fine-grained PATs (github_pat_*).
| Role | Agent Type | Use Case | |------|-----------|----------| | General | copilot | Broad research, code explanation, exploration | | Research | copilot-research | Research-focused exploration and analysis |
bash# Programmatic mode (non-interactive) copilot -p "<prompt>" --no-ask-user
When Copilot is active in a multi-provider workflow:
Providers:
🔴 Codex CLI - Implementation
🟡 Antigravity CLI - Security review
🟢 Copilot CLI - Research perspective
🔵 Claude - SynthesisIndicator legend:
The /octo:doctor providers check reports Copilot availability and auth method:
Providers:
✓ Copilot CLI installed (auth: keychain)When unauthenticated: ⚠ Copilot CLI installed but not authenticated When missing: ℹ Copilot CLI not installed (optional)
copilot -p prompt = 1 premium request from your monthly allowanceauto selector by default; pin OCTOPUS_COPILOT_MODEL to pass an explicit --model value🐙 **CLAUDE OCTOPUS ACTIVATED** - Multi-provider research mode
🔍 Discover Phase: Researching WebSocket authentication patterns
Providers:
🔴 Codex CLI - Technical implementation analysis
🟡 Antigravity CLI - Ecosystem research
🟢 Copilot CLI - Research perspective
🔵 Claude - Strategic synthesis🐙 **CLAUDE OCTOPUS ACTIVATED** - Multi-provider research mode
🔍 Discover Phase: Researching WebSocket authentication patterns
Providers:
🔴 Codex CLI - Technical implementation analysis
🟡 Antigravity CLI - Ecosystem research
🔵 Claude - Strategic synthesisWhen Copilot is not detected, it is silently omitted from the provider list.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | fail→pass | 23,446 | 6,673 | -72% | 1 | 1 | 0% | 3,202 | 1,378 | -57% | 0 | 0 | — |
case-01 | fail→pass | 13,232 | 8,062 | -39% | 1 | 1 | 0% | 1,470 | 1,710 | +16% | 0 | 0 | — |
case-02 | fail→fail | 22,970 | 9,036 | -61% | 1 | 1 | 0% | 4,632 | 1,828 | -61% | 0 | 0 | — |
case-03 | fail→pass | 27,737 | 6,904 | -75% | 1 | 1 | 0% | 2,514 | 2,496 | -1% | 0 | 0 | — |
case-13 | fail→pass | 14,768 | 6,888 | -53% | 1 | 1 | 0% | 1,575 | 1,412 | -10% | 0 | 0 | — |
case-04 | pass→pass | 12,595 | 11,199 | -11% | 1 | 1 | 0% | 1,391 | 2,261 | +63% | 0 | 0 | — |
case-05 | pass→pass | 7,480 | 11,315 | +51% | 1 | 1 | 0% | 1,476 | 2,268 | +54% | 0 | 0 | — |
case-06 | pass→pass | 5,701 | 8,569 | +50% | 1 | 1 | 0% | 1,066 | 1,699 | +59% | 0 | 0 | — |
case-07 | pass→pass | 5,034 | 7,204 | +43% | 1 | 1 | 0% | 890 | 1,546 | +74% | 0 | 0 | — |
case-08 | pass→pass | 17,806 | 3,887 | -78% | 1 | 1 | 0% | 2,246 | 1,732 | -23% | 0 | 0 | — |
case-09 | fail→pass | 18,549 | 9,557 | -48% | 1 | 1 | 0% | 2,247 | 1,762 | -22% | 0 | 0 | — |
case-10 | fail→pass | 10,554 | 6,704 | -36% | 1 | 1 | 0% | 1,899 | 1,369 | -28% | 0 | 0 | — |
case-11 | fail→pass | 10,498 | 1,819 | -83% | 1 | 1 | 0% | 1,796 | 1,420 | -21% | 0 | 0 | — |
case-14 | fail→pass | 17,496 | 5,676 | -68% | 1 | 1 | 0% | 2,122 | 2,100 | -1% | 0 | 0 | — |
case-15 | pass→pass | 10,430 | 7,174 | -31% | 1 | 1 | 0% | 1,701 | 1,386 | -19% | 0 | 0 | — |
case-16 | pass→pass | 9,927 | 8,526 | -14% | 1 | 1 | 0% | 1,442 | 1,776 | +23% | 0 | 0 | — |
case-17 | fail→pass | 10,666 | 6,917 | -35% | 1 | 1 | 0% | 955 | 1,304 | +37% | 0 | 0 | — |
case-18 | pass→pass | 12,552 | 2,810 | -78% | 1 | 1 | 0% | 2,096 | 1,579 | -25% | 0 | 0 | — |
case-19 | pass→pass | 16,638 | 3,419 | -79% | 1 | 1 | 0% | 2,012 | 1,790 | -11% | 0 | 0 | — |
case-20 | fail→pass | 35,897 | 6,625 | -82% | 1 | 1 | 0% | 2,737 | 1,390 | -49% | 0 | 0 | — |
case-21 | fail→pass | 11,179 | 8,332 | -25% | 1 | 1 | 0% | 1,842 | 1,495 | -19% | 0 | 0 | — |
case-22 | fail→pass | 16,483 | 3,387 | -79% | 1 | 1 | 0% | 1,758 | 1,692 | -4% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +55 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.