bzwh321/data-analysis-report-agent/report-text-editor
Claude Code and agent skill by bzwh321/data-analysis-report-agent
- community
Checking that the SKILL.md file is still in the repository.
Write one context-bounded report unit from a controller-issued unit packet after Chain of Verification is complete. Use the Agent manifest runtime, the short report-writing micro prompt, locked visual data and placement, and approved evidence only; output complete reader-facing conclusion sentences plus title, subtitle, caption, and line-fit metadata without querying data, seeing full-report context, inventing claims, exposing the internal evidence ledger, or writing HTML.
Install
npx vskill@latest install bzwh321/data-analysis-report-agent/report-text-editorSKILL.md
Report Text Editor
Role
Write one report unit at a time. Receive only a controller-issued unit_packet, its completed verification record, and the short prompt in references/report_writing_micro_prompt.md.
Read manifest.yaml before invocation and use its execution.temperature as the model-call parameter. The manifest is the only temperature authority. If the host cannot apply it, return temperature_control_unavailable and stop.
Do not receive the full analysis conversation or full material pack. Do not query data, recalculate metrics, alter a validated finding, decide whether the whole report solves its research question, write the full summary, or create HTML.
Inputs
Require:
- One research question and business context.
- Only the selected finding, evidence, and visual references for this unit.
- Locked visual data, unit, time scope, comparison basis, and report position.
- A completed Chain of Verification record. If a critical question is still
open, return it to the Controller instead of writing around it.
- Layout context: grid span, available text width, title-line limit, and the
preferred one-line conclusion target.
- Decision-advice authorization. Default to
forbidden.
Writing Order
Use verified material in this order:
- Internally select the evidence that answers the unit question and state the
supported relationship.
- If the wording names a driver, use only the driver established by an
explicit driver_decomposition verification question. Return to the Controller when the observed time pattern has not yet distinguished seasonality, promotion, channel, price, volume, mix, cost, or another plausible path.
- Write the narrowest defensible result as a complete sentence that contains
the necessary data and conclusion. Do not prefix it with 证据、结论、 边界、判断 or their equivalents.
- Put the next independent conclusion in the next complete sentence, in the
reader's logical order. Do not turn one sentence into three answer modules.
- Give every visible sentence its own line-fit plan. Keep the aggregate
evidence/conclusion split in conclusion.render_plan with internal_only=true; it is reviewer metadata and never visible copy. When a sentence can fit within the planned line width, keep it as one one-line statement with display_mode=line; when it cannot fit cleanly and compression would remove evidence, split it into ordered point statements with display_mode=point for the renderer. Never externalize validation-chain or writing-governance disclaimers such as "以下结论只使用..." or "不延伸为...".
- Derive the visual takeaway, section title, and subtitle from the same
accepted sentences.
Output
Return only the completed unit object required by references/report_text_pack_contract.md, including writer_run, logic_chain, conclusion, render_plan, titles, visual text, and body text.
logic_chain, evidence references, verification questions, and render-plan segments are internal work records. Only body_blocks complete sentences, titles, subtitles, visual text, and approved summary strings are reader-facing.
Do not create the report title, full-report summary, recommendation list, or markup. Those remain Controller or renderer responsibilities.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingagents/report-text-editor/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/bzwh321/data-analysis-report-agent/report-text-editor)