菜单
Cross-session learnings lifecycle — schema, storage, retrieval, and pruning of lessons learned during orchestrator sessions. Use when managing learnings via the /learn command.
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。
基于 SOC 职业分类
Demo pipeline state machine — 7-phase autonomous sequence with delegation context templates, phase transition logic, BLOCKED recovery strategies, and demo-state.json schema. Used exclusively by demo-conductor.
Cinematic narration style guide for demo-conductor — ANSI-coloured banner formats, live pipeline scoreboard, audience-facing language, phase summaries, and error narration patterns. Keeps the demo presentation-quality throughout.
Standardized completion and escalation protocol for subagent responses. Ensures the conductor can machine-parse every subagent return. Use when reporting completion status back to the orchestrator.
Agent Teams assembly and task injection — selects appropriate team, validates prerequisites, estimates cost, injects tasks into the shared task list, and manages team lifecycle.
Token and cost tracking with model tier enforcement
Complexity-based task routing with model tier selection
| name | learnings-mgmt |
| description | Cross-session learnings lifecycle — schema, storage, retrieval, and pruning of lessons learned during orchestrator sessions. Use when managing learnings via the /learn command. |
| user-invocable | false |
Learnings persist to artifacts/memory/learnings.jsonl — one JSON object per line.
{
"id": "learn-<timestamp-ms>",
"timestamp": "2026-03-30T12:00:00.000Z",
"category": "pattern | pitfall | convention | decision | tooling",
"note": "Free-text lesson learned",
"source": "session | user | review | incident",
"deleted": false
}
| Field | Required | Description |
|---|---|---|
id | Yes | learn- prefix + Unix epoch milliseconds — unique per entry |
timestamp | Yes | ISO 8601 |
category | Yes | One of: pattern, pitfall, convention, decision, tooling |
note | Yes | The lesson text — keep under 200 words |
source | Yes | One of: session, user, review, incident |
deleted | No | Defaults to false. Set true on soft-delete. |
Append a new JSON line to learnings.jsonl. Auto-assign id and timestamp. Default source to session unless the user specifies otherwise.
Read learnings.jsonl, filter out deleted: true entries, display the last 20 in reverse chronological order. Show: id, timestamp (date only), category, and truncated note (first 80 chars).
Read all non-deleted entries, filter where note contains the query string (case-insensitive). Display matching entries with full note text.
Find the entry by id. Set deleted: true and rewrite the line in place. Do not physically delete — this preserves audit history.
Read all non-deleted entries. Write a Markdown summary to artifacts/memory/learnings-export.md grouped by category. Include timestamp and full note text.
Agents may suggest adding a learning when:
The agent should present the suggestion and let the user confirm via /learn add <note>.
Over time, learnings accumulate. Recommend pruning when learnings.jsonl exceeds 200 entries:
artifacts/memory/learnings-archive.jsonl