メニュー
Collaboratively refine a high-level ambition into a set of chunk prompts.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | narrative-create |
| description | Collaboratively refine a high-level ambition into a set of chunk prompts. |
The operator wants to collaboratively develop this concept with you:
$ARGUMENTS
Create a short name handle that describes this concept. A short name should be 32 characters or less and words should be underscore separated. We will refer to this shortname later as .
Run ve narrative create <shortname> and note the created path. The narrative
will be created in docs/narratives/. Example output:
Created docs/narratives/<shortname>
We will refer to this path later as <narrative_path>.
Complete the template in <narrative_path>/OVERVIEW.md with the
information supplied by the operator and through further clarification
interactions with the operator. Completing the template includes writing
the proposed_chunks frontmatter array — populate it now with a prompt
entry for each chunk identified during refinement (set chunk_directory: null for each; /chunk-create will fill that in when the chunk is reified).
When populating proposed_chunks, understand the depends_on semantics.
Each entry in proposed_chunks can optionally declare a depends_on field.
This field has three meaningful states—the same null vs empty distinction
used in chunk GOAL.md files:
Omit the field (or set to null): You don't know this prompt's dependencies
yet. At chunk-create time, the orchestrator's conflict oracle will analyze.
Use depends_on: []: You explicitly know this prompt has no dependencies
on other prompts in this narrative. This bypasses oracle consultation.
Use depends_on: [0, 2]: This prompt depends on prompts at indices 0 and 2
in the same proposed_chunks array. At chunk-create time, these indices are
translated to chunk directory names.
Practical guidance:
depends_on to let
the oracle detect any subtle dependencies you might have missed.depends_on: [1] for chunk 3's entry.[] to assert this,
but omitting is safer when uncertain.See the OVERVIEW.md template's PROPOSED_CHUNKS section for the full semantics table.
Migrate the project's ACTIVE chunks to the present-tense, intent-owning standard. Audits each chunk's goal against the code it claims to govern; rewrites retrospective framing inline; logs over-claims that need operator triage; historicalizes chunks with no enduring intent. Designed for full-corpus migrations — fans out across many parallel sub-agents at 5 chunks per agent.
Create a new chunk of work and refine its goal. Use when the operator wants to start new intent-bearing work, chunk something, define a piece of work, or break work into a chunk.
Update code references in the current chunk and move both the PLAN.md and the GOAL.md to the ACTIVE state.
Wake an entity by loading its identity, memories, and operational context
Set up a project steward via interactive interview
Run the sleep cycle for an entity — extract memories and consolidate