omcustom:goal

Disciplined goal-to-execution workflow for any user task. Parses objective, asks only for materially missing requirements via ambiguity-gate, inspects repo via idea, plans via sdd-dev or deep-plan, ex

SKILL.md

GitHub

/goal — Disciplined Goal-to-Execution Workflow

Apply project discipline (planning, verification, cleanup, review) to any user-provided task. Acts as a thin orchestrator that routes to existing oh-my-customcode skills rather than duplicating logic.

Usage

/goal <task description>

Example: /goal Add a /weekly-summary command that aggregates last 7 days of git activity

Workflow

  1. Parse objective: Treat <task> as the concrete goal. Do NOT expand scope beyond what was asked.
  2. Gap detection: Invoke ambiguity-gate skill. Ask only for materially missing or risky requirements — no over-clarification on details that can be inferred.
  3. Repo inspection: Invoke idea skill (NL-intent codebase analysis) to identify relevant files, dependencies, and existing assets.
  4. Plan generation (scope-dependent):
    • Trivial scope (single file, <30 LOC, no API design): proceed with brief inline plan.
    • Non-trivial scope: invoke sdd-dev or deep-plan for research-validated planning.
  5. Safe execution: Follow R001 (safety), R009 (parallel), R010 (delegation). Delegate ALL file modifications to specialist agents.
  6. Completion verification: Apply R020 — verify ACTUAL outcome before declaring [Done]. Run task-type-specific checks (lint, build, test).
  7. Report: Output changed files, verification evidence, and remaining risks.

Output Format

[Goal] {original task verbatim}
[Plan] {brief plan or reference to deep-plan/sdd-dev output}
[Changes] {file list with paths}
[Verified] {evidence — exit codes, test counts, build output}
[Risks] {remaining concerns or "none"}

When to Use

  • Generic multi-step task without obvious specialist routing
  • User wants a single, memorable entry point for "do this task with full discipline"
  • Work that benefits from planning + verification + structured reporting

When NOT to Use

Scenario Use Instead
Pure research/analysis research, idea, memory-recall
Specialist domain task Routing skills (dev-lead-routing, de-lead-routing, qa-lead-routing)
Trivial single-file edit Delegate directly to specialist agent
Bug debugging superpowers:systematic-debugging

Integration with Existing Skills

Stage Skill Role
Gap detection ambiguity-gate Score request clarity, ask only critical gaps
Repo inspection idea NL-intent codebase analysis
Planning (research-validated) deep-plan research → plan → verify cycle
Planning (SDD discipline) sdd-dev Stage gates with planning-first workflow
Verification superpowers:verification-before-completion Evidence before assertions
Reporting result-aggregation Concise output composition

Design Notes

This skill is intentionally a thin wrapper / orchestrator. It does NOT duplicate planning, gap-detection, or verification logic. Its value:

  • Single memorable command for the disciplined workflow
  • Stable entry point that routes to whichever underlying skills are best at the moment
  • Documented contract that callers can rely on regardless of internal skill evolution

If a future skill replaces idea or sdd-dev, only the integration table updates — /goal remains stable.

Cross-References

  • Issue: #1109 — Port /goal workflow to Claude Code skill surface
  • Rules: R001 (safety), R003 (response style), R010 (delegation), R020 (completion verification)
  • Original Codex /goal workflow this ports from: discipline-first goal execution