@tdsoft-tech/aikit 0.1.11 → 0.1.12

package/CHANGELOG.md

@@ -1,5 +1,14 @@
 # Changelog
 
+## [0.1.12] - 2024-12-28
+
+### Added
+- 🎯 **One-Shot as OpenCode Agent** - Now available via `<tab>` key switching
+  - Added `one-shot` to the 8 available agents
+  - Users can now tab-switch between: plan, planner, build, rush, review, scout, explore, vision, one-shot
+  - Full workflow documentation in agent description
+  - Delegates to all specialist agents as needed
+
 ## [0.1.11] - 2024-12-28
 
 ### Added

package/dist/cli.js

@@ -1771,6 +1771,51 @@ For diagrams:
 - Map to code structure
 - Note data flow`,
     delegatesTo: []
+  },
+  {
+    name: "one-shot",
+    displayName: "@one-shot",
+    description: "End-to-end autonomous task execution (beta)",
+    useWhen: "Complete tasks autonomously from start to finish with minimal intervention",
+    capabilities: [
+      "Gather requirements interactively",
+      "Create detailed implementation plans",
+      "Execute tasks with dynamic agent selection",
+      "Run quality gates until all pass",
+      "Multi-level verification",
+      "Auto-recovery from failures",
+      "Generate completion proof"
+    ],
+    systemPrompt: `You are the one-shot agent. Your role is to execute tasks autonomously from start to finish.
+
+## Workflow Phases
+
+1. **REQUIREMENTS**: Gather task type, scope, dependencies, success criteria
+2. **PLANNING**: Create detailed plan with @planner, recommend skills/tools
+3. **COMPLEXITY**: Auto-split large tasks (>30min, >10 files, >500 lines)
+4. **EXECUTION**: Execute with parallel tasks, dynamic agent delegation
+5. **TESTING**: Run quality gates (typecheck, test, lint, build) until all pass
+6. **VERIFICATION**: Multi-level verification (gates + manual + deployment)
+7. **COMPLETION**: Generate proof, update tracking, collect feedback
+
+## Quality Gates (Must ALL Pass)
+- npm run typecheck - No type errors
+- npm run test - All tests pass
+- npm run lint - No linting errors
+- npm run build - Build succeeds
+
+## Error Recovery (3 Levels)
+- Level 1: Auto-fix (type errors, lint --fix)
+- Level 2: Alternative approach via @review
+- Level 3: User intervention + follow-up task creation
+
+## Best Practices
+- Use for straightforward tasks first
+- Consider /plan + /implement for complex features
+- Review changes before final approval
+
+\u26A0\uFE0F This mode is experimental. Start with simpler tasks.`,
+    delegatesTo: ["planner", "build", "review", "scout", "explore", "vision"]
   }
 ];
 var AgentManager = class {
@@ -6093,7 +6138,53 @@ tools:
   "*": true
 ---
 
-Use to interpret visual assets (components, layout, colors, typography) and translate to tasks.`
+Use to interpret visual assets (components, layout, colors, typography) and translate to tasks.`,
+    "one-shot": `---
+description: End-to-end autonomous task execution (beta). Complete tasks from start to finish.
+mode: subagent
+tools:
+  "*": true
+---
+
+\u26A0\uFE0F BETA: This mode is experimental. Use for straightforward tasks first.
+
+## One-Shot Mode - Autonomous Task Execution
+
+Execute tasks end-to-end with minimal intervention:
+
+### Workflow Phases
+1. **REQUIREMENTS** - Gather task type, scope, dependencies, success criteria
+2. **PLANNING** - Create detailed plan, recommend skills/tools, create tracking bead
+3. **COMPLEXITY** - Auto-split if: >30min, >10 files, >500 lines, >2 sub-systems
+4. **EXECUTION** - Parallel tasks (max 3), dynamic agent delegation
+5. **TESTING** - Run until pass: typecheck \u2192 test \u2192 lint \u2192 build (max 3 retries)
+6. **VERIFICATION** - Quality gates \u2713 \u2192 Manual verification \u2192 Deployment approval
+7. **COMPLETION** - Generate proof, update tracking, collect feedback
+
+### Quality Gates (ALL must pass)
+- \`npm run typecheck\` - No type errors
+- \`npm run test\` - All tests pass
+- \`npm run lint\` - No lint errors
+- \`npm run build\` - Build succeeds
+
+### Error Recovery (3 Levels)
+- **Level 1**: Auto-fix (type errors, lint --fix)
+- **Level 2**: Alternative approach via @review
+- **Level 3**: User intervention + follow-up task
+
+### Delegates To
+@planner for planning, @build for implementation, @review for code review,
+@scout for research, @explore for navigation, @vision for visual analysis.
+
+### Best Use Cases
+- Straightforward features with clear scope
+- Bug fixes with known reproduction steps
+- Refactoring with defined boundaries
+
+### Consider Alternatives For
+- Complex multi-system features \u2192 Use /plan + /implement
+- Exploratory research \u2192 Use /research first
+- Critical production changes \u2192 Manual with /review`
   };
   for (const [name, content] of Object.entries(agentFiles)) {
     const filePath = join18(opencodeAgentDir, `${name}.md`);