feliperyba
feliperyba/ralph-orchestra
相关职业SOC
基于 SOC 职业分类
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/feliperyba/ralph-orchestra --skill pm-organization-scale-adaptive命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
Complete Developer workflow orchestration - task research sequence, implementation flow, validation gates, PRD synchronization, exit conditions.
Complete Game Designer workflow - skill invocation protocol, GDD creation, playtest flow with GDD review, design sessions. MUST load before starting assignments.
Complete PM Coordinator workflow - task assignment, project orchestration, PRD management, worker coordination. Use proactively when starting PM agent work.
正在显示 SKILL.md
| name | pm-organization-scale-adaptive |
| description | Adjust planning depth and agent behavior based on PRD complexity level |
| category | organization |
"Right-size the process – small projects need less ceremony, large projects need more structure."
Use at:
// Determine project scale (reads both PRD files)
const prd = readJson("prd.json");
const backlog = readJson(prd.backlogFile || "prd_backlog.json");
const allItems = [...prd.items, ...backlog.backlogItems];
const taskCount = allItems.filter((i) => !i.passes).length;
const scale =
taskCount <= 3 ? 0 : taskCount <= 8 ? : taskCount <= ? : taskCount <= ? : ;
config = [scale];
Note: Tasks are split between prd.json (active queue: 5 tasks) and prd_backlog.json (backlog: ~70 tasks). PM must read both files to get accurate task counts for scale detection.
| Scale | Task Count | Process Depth | Retrospective | Skill Updates |
|---|---|---|---|---|
| 0 - Micro | 1-3 | Minimal | Brief notes | None |
| 1 - Small | 4-8 | Light | Quick review | Anti-patterns only |
| 2 - Medium | 9-15 | Standard | Full process | Update relevant |
| 3 - Large | 16-30 | Comprehensive | Deep analysis | Create new skills |
| 4 - Enterprise | 31+ | Full methodology | Multi-phase | Full skill suite |
Minimal ceremony for tiny projects:
## Process Adjustments
- Skip formal retrospective file
- Use inline progress notes
- No skill improvement phase
- Simple task → implement → validate → done
## State Files
- prd.json.session (minimal session state)
- prd.json.items[{taskId}] (task details)
- Skip: retrospective.txt, handoff-log.json
Light process for small projects:
## Process Adjustments
- Brief retrospective (inline in progress.txt)
- Quick learnings capture
- Anti-pattern documentation only
- Single-pass validation
## Retrospective Format (Inline)
### [TIMESTAMP] Task Complete: {{TASK_ID}}
- Done: {{what}}
- Learned: {{key insight}}
- Watch out: {{anti-pattern if any}}
Standard process for typical projects:
## Process Adjustments
- Full file-based retrospective
- Agent contributions required
- Skill updates for gaps
- Risk identification
## Full Process
1. Create retrospective.txt
2. Wait for Developer + QA input
3. PM synthesis
4. Update skills if gaps found
5. Proceed to next task
Comprehensive process for complex projects:
## Process Adjustments
- Deep retrospective analysis
- Milestone reviews every 5 tasks
- Create new skill files
- Reference folder documentation
- Risk heat map tracking
## Additional Steps
1. Milestone review at 25%, 50%, 75%
2. Create domain-specific skill files
3. Build reference documentation
4. Track technical debt
5. Velocity monitoring
Full methodology for large projects:
## Process Adjustments
- Multi-phase planning
- Sprint boundaries
- Full skill suite creation
- Architectural decision records
- Quality gates
## Enterprise Features
1. Sprint planning (5-8 tasks per sprint)
2. Architectural retrospectives
3. Full skill documentation
4. Integration testing phases
5. Release milestones
function detectScale(prd, backlog) {
// Read both files for complete task picture
const prd = readJson("prd.json");
const backlog = readJson(prd.backlogFile || "prd_backlog.json");
const allItems = [...prd.items, ...backlog.backlogItems];
const remaining = allItems.filter((i) => !i.passes).length;
const total = allItems.length;
const complexity = allItems.reduce(
(sum, i) => sum + (i.category === 'architectural' ? 3 : i.category === 'integration' ? 2 : 1),
0
);
// Base scale on task count
let scale =
remaining <= 3 ? 0 : remaining <= 8 ? 1 : remaining <= 15 ? 2 : remaining <= 30 ? 3 : 4;
// Adjust for complexity
if (complexity / total > 2) scale = Math.min(4, scale + 1);
// Adjust for dependencies
const depCount = allItems.reduce((sum, i) => sum + i.dependencies.length, 0);
if (depCount / total > 1.5) scale = Math.min(4, scale + 1);
return scale;
}
❌ DON'T:
✅ DO:
At session start:
At each retrospective: