本文へ移動
cccskills
無料GitHub で公開

chatcrystal-task-writeback

Write reusable ChatCrystal task memories after substantive work completes. Use when implementation or debugging produced a durable fix, pitfall, pattern, or decision worth preserving, and when the environment can either persist it through `write_task_memory` or emit a structured memory candidate for later save.

インストール方法を見る

含まれるファイル(2)

  • SKILL.md4.4 KB
  • agents/openai.yaml237 B

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

ChatCrystal Task Writeback

Use this skill after meaningful work completes. Let ChatCrystal Core decide whether a memory should be skipped, merged, or created.

Workflow

  1. Trigger only after substantive work completes and the result produced reusable knowledge.
  2. Good candidates include:
    • a durable fix
    • a meaningful pitfall
    • a reusable pattern
    • an explicit engineering decision

Quality Gate Preflight

Before calling write_task_memory, verify the candidate is an experience asset, not a work log.

Do not call write_task_memory when:

  • The summary only says what changed, without why it mattered.
  • There is no clear problem, decision, pitfall, resolution, or reusable pattern.
  • The task produced no durable lesson for future work.

Required field shape by memory.outcome_type:

  • fix: include root_cause or resolution.
  • pitfall: include pitfalls plus resolution or reusable_patterns.
  • decision: include decisions.
  • pattern: include reusable_patterns.

Write memory.summary as a durable experience conclusion, not as "I did X". Automatic writeback must include source_run_key; without it, emit a structured candidate for later save instead of silently writing manual memory.

  1. If write_task_memory is available and the runtime has a stable run or session key, call it with:
    • mode: "auto"
    • source_run_key: the stable run or session key
    • task.goal, task.task_kind, task.project_dir, task.cwd, task.branch
    • task.source_agent when known
    • memory.summary: concise and durable, not chatty
    • memory.outcome_type: one of pitfall, fix, pattern, decision
    • memory.root_cause, memory.resolution, memory.error_signatures, memory.files_touched, memory.tags when relevant
  2. Use mode: "manual" only for an explicit manual save flow, never as a silent fallback for automatic writeback.
  3. Default to project-scoped knowledge. Only use global scope for clearly cross-project knowledge and only in an explicit manual save flow.
  4. Report the persistence result accurately:
    • created
    • merged
    • skipped

Example MCP Input

Use this shape when calling write_task_memory after a completed task with a stable run key:

{
  "mode": "auto",
  "source_run_key": "agent-session-or-run-id",
  "scope": "project",
  "task": {
    "goal": "Fix MCP recall schema validation",
    "task_kind": "debug",
    "project_dir": "/path/to/project",
    "cwd": "/path/to/project",
    "branch": "fix/mcp-recall-schema",
    "files_touched": [
      "server/src/services/memory/schemas.ts",
      "server/src/cli/mcp/server.ts"
    ],
    "error_signatures": [
      "invalid_type",
      "recall_for_task schema mismatch"
    ],
    "source_agent": "codex"
  },
  "memory": {
    "summary": "MCP tool schemas must stay aligned with memory service request shapes; mismatches surface as validation failures before tool execution.",
    "outcome_type": "fix",
    "root_cause": "The MCP tool input shape diverged from the service schema.",
    "resolution": "Reused the shared request shape in the MCP server registration.",
    "files_touched": [
      "server/src/services/memory/schemas.ts",
      "server/src/cli/mcp/server.ts"
    ],
    "tags": [
      "mcp",
      "memory-loop",
      "schema"
    ]
  }
}

Full Mode

Full mode requires ChatCrystal Core plus MCP access to write_task_memory.

  • Do not reimplement skip or merge logic in the skill.
  • Trust Core to decide whether the memory should be skipped, merged, or created.
  • If the tool reports skipped, keep the task result but do not force persistence.

Degraded Mode

If ChatCrystal Core or the MCP tool is unavailable, or if no stable run key exists for automatic writeback:

  • Do not claim that the memory was saved.
  • Emit a structured memory candidate in the response for later save.
  • Keep the candidate aligned with the write_task_memory.memory payload shape so it can be reused later.
  • Do not instruct the runtime to auto-install ChatCrystal Core as part of the skill flow.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Use Cursor's browser aria snapshots to audit a page for accessibility issues — missing labels, broken tab order, contrast, and ARIA misuse.

日本語の概要は準備中です。原文の説明を表示しています。

0xAidan/polymarket-bot-test42026年9月4日 更新

Add PostHog analytics to a web application, including event tracking, page views, feature flags, and session replay.

日本語の概要は準備中です。原文の説明を表示しています。

0xAidan/polymarket-bot-test42026年9月4日 更新

Generate OpenAPI/Swagger documentation for an API, including endpoint schemas, request/response types, and interactive docs UI.

日本語の概要は準備中です。原文の説明を表示しています。

0xAidan/polymarket-bot-test42026年9月4日 更新

Add authentication to a web application using NextAuth.js (Auth.js), including OAuth providers, session management, and protected routes.

日本語の概要は準備中です。原文の説明を表示しています。

0xAidan/polymarket-bot-test42026年9月4日 更新

Dockerize an application with a production-ready Dockerfile, docker-compose setup, and .dockerignore.

日本語の概要は準備中です。原文の説明を表示しています。

0xAidan/polymarket-bot-test42026年9月4日 更新

Set up Playwright end-to-end testing in a project, including test configuration, example tests, and CI integration.

日本語の概要は準備中です。原文の説明を表示しています。

0xAidan/polymarket-bot-test42026年9月4日 更新

0xAidan のスキルをすべて見る

このスキルの問題を報告する