ace-analysis
Analyzes sessions using ACE framework to generate prompt improvement suggestions focused on durable patterns.
ソース情報
- リポジトリ
- digi4care/opencode-mastery
- ソースの最終更新活動
- 2026年2月21日 09:55
- 検出された SKILL.md の言語
- 英語
- スター
- 0
- フォーク
- 0
インストール方法
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
ソースファイルを確認
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
ファイルエクスプローラー
3 ファイルSKILL.md を表示中
SKILL.md
ソースの指示 · 読み取り専用プレビュー- 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
```
GitHubで見る