claude-dailylisted
Install: claude install-skill longyunfeigu/claude-daily
<!--
Schema influences:
- US Army AAR (Intent / Reality / Gap / Adapt)
- Argyris double-loop learning (single-loop action vs double-loop assumption)
- Episodic→semantic memory consolidation (rules over titles, abstraction over recap)
These are scaffolding for the schema, not vocabulary to quote in output.
-->
# Daily Session Report
This skill turns local Claude Code transcripts into PRD-compliant payloads
for the Compound Daily ingestion pipeline. This file defines how Claude Code
drives the run end-to-end; `references/report-prompts.md` is the source of
truth for Session Card, Daily Report, evidence, and audience rules.
## Trigger
When the user says one of:
- "跑 claude-daily" / "生成今天的日报" / "做今日复盘" / "daily report"
→ Run all 6 steps below (default; includes upload).
- "跑 claude-daily dry-run" / "只看不发" / "preview" / "先看看不发"
→ Run steps 1-5 only; stop after emit shows the summary table.
- "跑昨天的" / explicit "--date YYYY-MM-DD"
→ Use specified date instead of today.
- "重新发一遍" / "force"
→ Pass `--force` to upload (re-uploads regardless of existing acks).
- "跑 claude-daily ai-assessment" / "生成 AI 使用能力评估" / "生成管理者版 AI 使用评估"
→ Run the **AI 使用评估模式** below, NOT the 6-step daily pipeline. 单人;
单日用 `--date`,多日窗口用 `--from/--to`(优先建议多日窗口——单日样本
支撑不了「当前水平」结论)。默认含 upload,dry-run / "只看不发" 跳过。
If the user's intent is ambiguous (e.g. "今天看一下"), ask once before
starting. Default the mode slot to dry-run when ambiguous, not auto-upload.
## 6-Step Pipeline
### 1. Prepare cont