$npx -y skills add jmstar85/oh-my-githubcopilot --skill omg-autopilotFull autonomous execution from idea to working code. Activate when user says: autopilot, auto-pilot, autonomous, build me, create me, make me, full auto, handle it all, or "I want a/an...
| 1 | # OMG Autopilot |
| 2 | |
| 3 | OMG Autopilot takes a brief product idea and autonomously handles the full lifecycle: requirements analysis, technical design, planning, parallel implementation, QA cycling, and multi-perspective validation. |
| 4 | |
| 5 | ## When to Use |
| 6 | - User wants end-to-end autonomous execution from an idea to working code |
| 7 | - Task requires multiple phases: planning, coding, testing, and validation |
| 8 | - User wants hands-off execution |
| 9 | |
| 10 | ## When NOT to Use |
| 11 | - User wants to explore options → use `/plan` |
| 12 | - Single focused code change → use `/ralph` or delegate to @executor |
| 13 | - Quick fix or small bug → delegate directly to @executor |
| 14 | |
| 15 | ## Interactive Hook Protocol |
| 16 | |
| 17 | Use `vscode_askQuestions` at critical decision gates during autopilot execution (when available). |
| 18 | If `vscode_askQuestions` is NOT available (e.g., Copilot CLI), present numbered options in markdown. |
| 19 | Autopilot is designed to be mostly autonomous, but hooks fire at ambiguity points and failure recovery. |
| 20 | |
| 21 | ### When to Fire Hooks |
| 22 | | Trigger Point | Question Type | |
| 23 | |---------------|---------------| |
| 24 | | Phase 0: Vague input | Redirect to deep-interview or proceed with assumptions | |
| 25 | | Phase 0: Spec review | Confirm extracted spec before planning | |
| 26 | | Phase 3: Repeated QA failure | Ask user for direction on persistent errors | |
| 27 | | Phase 4: Validation rejection | Surface reviewer concerns for user decision | |
| 28 | |
| 29 | ## Execution Pipeline |
| 30 | |
| 31 | ### Phase 0 - Expansion |
| 32 | Turn the user's idea into a detailed spec. |
| 33 | - If ralplan consensus plan exists (`.omg/plans/ralplan-*.md`): Skip Phase 0 AND Phase 1 → jump to Phase 2 |
| 34 | - If deep-interview spec exists (`.omg/specs/deep-interview-*.md`): Use pre-validated spec, skip to Phase 1 |
| 35 | - If input is vague: **HOOK** via `vscode_askQuestions`: |
| 36 | ``` |
| 37 | header: "autopilot-vague-input" |
| 38 | question: "Your request is broad. How should I proceed?" |
| 39 | options: [ |
| 40 | { label: "Run /deep-interview first", description: "Thorough requirements gathering", recommended: true }, |
| 41 | { label: "Proceed with my best interpretation", description: "I'll extract requirements and show you for approval" }, |
| 42 | { label: "Let me clarify — here's more detail..." } |
| 43 | ] |
| 44 | allowFreeformInput: true |
| 45 | ``` |
| 46 | - Otherwise: @analyst extracts requirements, @architect creates technical specification |
| 47 | - **HOOK: Spec confirmation** before proceeding to planning: |
| 48 | ``` |
| 49 | header: "autopilot-spec-review" |
| 50 | question: "Here's what I understood: [1-line summary]. Proceed?" |
| 51 | options: [ |
| 52 | { label: "Looks good — proceed", recommended: true }, |
| 53 | { label: "Mostly right, but adjust..." }, |
| 54 | { label: "Wrong direction — let me re-explain" } |
| 55 | ] |
| 56 | allowFreeformInput: true |
| 57 | ``` |
| 58 | - Output: `.omg/autopilot/spec.md` |
| 59 | - Track phase: `omg_write_state(phase="expansion_done")` |
| 60 | |
| 61 | ### Phase 1 - Planning |
| 62 | Create an implementation plan from the spec. |
| 63 | - @architect creates plan (direct mode) |
| 64 | - @critic validates plan |
| 65 | - Output: `.omg/plans/autopilot-impl.md` |
| 66 | - Track: `omg_write_state(phase="planning_done")` |
| 67 | |
| 68 | ### Phase 2 - Execution |
| 69 | Implement the plan using parallel execution. |
| 70 | - Route tasks by complexity to @executor |
| 71 | - Run independent tasks in parallel |
| 72 | - Track: `omg_write_state(phase="execution_done")` |
| 73 | |
| 74 | ### Phase 3 - QA |
| 75 | Cycle until all tests pass (max 5 cycles). |
| 76 | - Build, lint, test, fix failures |
| 77 | - Stop early if the same error repeats 3 times (fundamental issue) |
| 78 | - **HOOK: On 2nd repeat of same failure** via `vscode_askQuestions`: |
| 79 | ``` |
| 80 | header: "autopilot-qa-stuck" |
| 81 | question: "Same error persists after [N] cycles: [error summary]. How to proceed?" |
| 82 | options: [ |
| 83 | { label: "Try a different approach", recommended: true }, |
| 84 | { label: "Skip this check — it's acceptable" }, |
| 85 | { label: "I know the fix — let me explain..." }, |
| 86 | { label: "Abort autopilot" } |
| 87 | ] |
| 88 | allowFreeformInput: true |
| 89 | ``` |
| 90 | - Track: `omg_write_state(phase="qa_done")` |
| 91 | |
| 92 | ### Phase 4 - Validation |
| 93 | Multi-perspective review in parallel. |
| 94 | - @architect: Functional completeness |
| 95 | - @security-reviewer: Vulnerability check |
| 96 | - @code-reviewer: Quality review |
| 97 | - All must approve; on rejection: |
| 98 | **HOOK** via `vscode_askQuestions`: |
| 99 | ``` |
| 100 | header: "autopilot-validation-rejection" |
| 101 | question: "[Reviewer] flagged: [issue summary]. Action?" |
| 102 | options: [ |
| 103 | { label: "Fix the issues", recommended: true }, |
| 104 | { label: "Accept risk — proceed anyway" }, |
| 105 | { label: "I need to review this myself first" } |
| 106 | ] |
| 107 | allowFreeformInput: true |
| 108 | ``` |
| 109 | - Fix and re-validate on rejection |
| 110 | - Track: `omg_write_state(phase="validation_done")` |
| 111 | |
| 112 | ### Phase 5 - Cleanup |
| 113 | Delete all state files on successful completion. |
| 114 | - Run `/cancel` for clean exit |
| 115 | |
| 116 | ## Stop Conditions |
| 117 | - Same QA error persists across 3 cycles → report fundamental issue |
| 118 | - Validation keeps failing after 3 rounds → report issues |
| 119 | - User says "stop", "cancel", or "abor |