ace-analysis
Analyzes sessions using ACE framework to generate prompt improvement suggestions focused on durable patterns.
소스 정보
- 저장소
- digi4care/opencode-mastery
- 최근 소스 활동
- 2026년 2월 21일 09:55
- 감지된 SKILL.md 언어
- 영어
- 스타
- 0
- 포크
- 0
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
파일 탐색기
3 개 파일SKILL.md 표시 중
SKILL.md
소스 지침 · 읽기 전용 미리보기- name
- ace-analysis
- description
- Analyzes sessions using ACE framework to generate prompt improvement suggestions focused on durable patterns.
- license
- MIT
- compatibility
- opencode
- metadata
- {"author":"OpenCode Community","version":"1.0.0"}
# ACE Analysis Skill
Analyzes AI agent sessions using ACE (Agentic Context Engineering) framework.
## Purpose
Evaluate session quality and generate improvement suggestions for prompts, skills, and commands.
## Triggers
- `/ace-reflect` command
- Session analysis requests
- Prompt optimization tasks
## Workflow
### 1. Collect Session Data
**Preferred:** `om-session` plugin tools
```
session-list → get recent sessions
session-read → get messages from session
```
**Fallback:** OpenCode CLI
```bash
opencode session list -n 10 --format json
```
**Last resort:** Direct context analysis (current conversation)
### 2. Determine Analysis Mode
| Mode | Flag | Scope |
| --------- | ------------- | ----------------------------------- |
| Default | (none) | Text analysis, 2-3 findings |
| Verbose | `--verbose` | Deeper analysis, more findings |
| Technical | `--technical` | Include IDs, timestamps, tool stats |
| Full | `--full` | Complete analysis with all metadata |
### 3. Analyze Session Content
Focus on **patterns**, not individual mistakes:
- Response quality (clarity, accuracy, completeness)
- Tool usage efficiency (redundant calls, missing tools)
- Context management (pruning, distillation)
- Workflow adherence (approval gates, context loading)
### 4. Score Session
Use scoring rubric from `references/scoring-rubric.md`:
| Criterium | Weight | Focus |
| ------------ | ------ | ------------------------------------ |
| Completeness | 5 | All requested tasks completed |
| Accuracy | 5 | Correct solutions, no hallucinations |
| Efficiency | 5 | Minimal iterations, good tool usage |
| Clarity | 5 | Clear communication, good structure |
| Relevance | 5 | On-topic, no unnecessary tangents |
**Max Score:** 25
### 5. Generate Report
Use template from `references/report-template.md`:
- Summary (1-2 sentences)
- Scores table
- Findings (patterns, not individual mistakes)
- Suggestions (specific, actionable)
- Decision (no changes / review / changes recommended)
### 6. Return to Orchestrator
Return formal report for presentation to user.
## References
- `references/report-template.md` - Report structure
- `references/scoring-rubric.md` - Detailed scoring criteria
## Error Handling
| Situation | Action |
| -------------------- | -------------------------------- |
| No session data | Analyze current context directly |
| Plugin unavailable | Use CLI fallback |
| Subagent spawn fails | Analyze in current context |
## Configuration
```yaml
features:
ace:
enabled: true
max_subagent_depth: 2
auto_apply_suggestions: false
```
GitHub에서 보기