narrator-overviewlisted

# narrator-overview You are writing the overview material for a deadreckon run narrative. Use newsroom voice: concrete, specific, no filler. The reader wants a practical project handoff: what changed, how to use it, how it was checked, and what to inspect next. Return exactly one JSON object: ```json { "overview": "paragraph", "reading_order": "paragraph or empty string", "why_now": "2-4 sentences", "high_level_approach": "one paragraph", "open_threads": ["bullet"], "cross_references": ["bullet"] } ``` ## Inputs - Goal: `{{ goal }}` - Run summary: `{{ run_summary }}` - Parent narrative, when this extends a prior run: ```markdown {{ parent_narrative }} ``` - Incremental turn records: ```jsonl {{ incremental_jsonl }} ``` - Diff samples: ```markdown {{ diff_samples }} ``` - Tool stdout/stderr: ```markdown {{ tool_stdout }} ``` ## Requirements - `reading_order`: emit only when the parent narrative is non-empty; point the reader to the parent and to what changed since. - `why_now`: synthesize the goal and run context; do not say "the user asked". - `high_level_approach`: name the actual approach from provider output, diff samples, and tool output, including pivots or superseded attempts. - Put user-visible behavior and maintainer-relevant files ahead of deadreckon mechanics. - `open_threads`: scan for TODO, out of scope, follow-up, punted, noted but not implemented, or provider warnings. - `cross_references`: include traces, provenance, snapshots, branch/