# Reporting Skills

> Use when the user asks for an executive summary, technical report, or complete report. Provides specialized knowledge and formatting guidelines for generating high-quality reports in different formats.

- Skill: `tankygranny05/reporting-skills` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add tankygranny05/reporting-skills`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tankygranny05/reporting-skills/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: tankygranny05 (https://skillmd.com/u/tankygranny05)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/tankygranny05/reporting-skills

---

<!-- [Edited by Claude: 2a51fb4e-1a96-4d2f-a1e1-d5bb1e208b1f] -->

# Reporting Skills

[Created by Claude: cd8e659b-5005-4a0b-a6f5-1b4f732a1d69 2026-01-28]
*[Edited by Codex: 019c058b-abb6-7f31-ba69-da517fcadecb 2026-01-28]*

Use when the user asks for an **executive summary**, **technical report**, or **complete report**.

## Purpose

Generate investigation reports in different formats based on user preference.

---

## MANDATORY: Read Example Reports First

Before generating any report, you MUST read ALL THREE example files in the `examples/` folder:

1. `examples/1_technical_report.md` - **MUST READ** - Codex-style technical report (raw dumps, call traces, line numbers)
2. `examples/2_unformatted_executive_summary.md` - **MUST READ** - Shows what NOT to do (content correct, formatting wrong)
3. `examples/3_ideal_executive_summary.md` - **MUST READ** - Shows the IDEAL executive format

**Optional:**
4. `examples/4_style_differences.md` - Claude vs Codex style tradeoffs (low priority)

---

## User Preferences

- **Quantification**: The user prefers numerical scores on a 1-10 scale when evaluating quality, correctness, or completeness. Always provide scores when assessing the state of something.
- **Tables**: Use tables for structured data (scores, comparisons, key evidence) but keep them concise (2-4 rows typically)
- **Flat bullets**: Avoid deeply nested bullet points - flatten to single level when possible

---

## Report Types

- **"executive summary"** → Executive Report format only
- **"technical report"** → Technical Report format only
- **"complete report"** or **"complete reports"** → BOTH versions, clearly separated with different markdown headers:

```
## Executive Summary
[executive content here]

## Technical Report
[technical content here]
```

---

## Executive Report

### The Verbatim Prompt

> Can you tell me in simple terms what this means for our implementation? Be less event oriented, don't quote specific code terms unless I specifically ask for "proof" or how did you know this. Mention the pattern and evaluate the risk. Always link the information you're presenting towards our goal. This should be a slightly technical executive summary. You could mention core architectural artifacts, but don't go into details.

### Concept Budget

The user can only keep **max 4 item names** in their head. Use this budget very carefully - most of the time this means giving high-level modules, architecturally significant ones, or whatever is most related to the user's workflow.

### The Formatting Prompt (Verbatim)

> Good, Some improvement, however you should use \` and ** more, header should have slightly different formatting / color also, basically think of this as a human friendly md (don't put stuff in wrapped md block, just type things out like we're in a md file already
>
> example:
>
> ## Main Locations
>
> | Location | Purpose |
> |----------|---------|
> | `/Users/sotola/db23_folder/` | Primary data storage (~90+ GB of pickles and transaction data) |
> | `/Users/sotola/PycharmProjects/mac_local_m4/` | Pipeline Python scripts |
> | `/Users/sotola/PycharmProjects/db23/` | Excel splitting utilities |
>
> ## Available Skills
>
> You have three skills for working with db23:
> - **`db23-pipeline`** – Run the 6-step data ingestion pipeline
> - **`db23-excel-split`** – Split multi-day Excel files into single-day files
> - **`db23-create-day-project`** – Create new Viber Stock OCR project folders
>
> ## Key Files
>
> - `df_trans_numerical.pickle` (7.1 GB) – Master numerical transaction dataframe
>
> now, try again

---

## Technical Report

### Requirements

- Be exact with file paths and line numbers
- Give assertions that are **cheap to verify** – an automated test script should report true/false immediately
- When codebase relationships change, verification should detect it instantly

### Style (Codex-style: agent/code-native)

- **Minimal markdown** - use **bold** for section labels, not `##` headers
- **Call chain traces** - show path to chokepoint: `Qf()` → `stop-hook runner` (`cli.js:456240`) → terminal paths
- **Dump technical terms** - `__cc_emitMainStopHookAndTurnEnd`, `cli.js:456313` - agents parse these natively
- **Arrow notation** - `A()` → `B()` → `C()`, `agentId && !__cc_isMainAgent(agentId)` → reject
- **Line numbers inline** - `cli.js:401530`, `:401554`, `:401580`
- **Verify commands** - `rg -c "pattern" file` → expected count

See `examples/1_technical_report.md` (MUST read).

For style comparison, see `examples/4_style_differences.md` (optional).