# Business Case Study Public

> Write the tier-3 public case study — hard sanitization, every headline claim triangulated against external evidence, candid limitations mandatory

- Skill: `tony/business-case-study-public` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add tony/business-case-study-public`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tony/business-case-study-public/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: tony (https://skillmd.com/u/tony)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/tony/business-case-study-public

---


# Public Case Study (tier 3)

Turn a run package into a case study fit for external publication.
Tier 3 is the hard sanitization tier: no internal identifiers of any
kind, aggregates and ranges only, every headline claim triangulated.
Value is engineer-hours, cycle time, and capacity; never money.

Read `references/interim-format.md` first — it
defines run location and the completeness gate. Defer these rendering
references until the gate passes; a refused render never loads them:

- `references/provenance.md` — tag rendering,
  anti-inflation rules, the no-currency contract.
- `references/evidence.md` — the external
  evidence table and the triangulation procedure.
- `references/audiences.md` — the tier-3
  contract and the final checklist.

User arguments: $ARGUMENTS

## Context

Recent runs — run this command and read the output:

```bash
sh -c 'ls -dt "$HOME"/Documents/*/business/ /mnt/c/Users/*/Documents/*/business/ 2>/dev/null' | head -5 | grep . || echo "(no runs found)"
```

## Procedure

### 1. Locate the run

Use the `$ARGUMENTS` path if given, else the newest run per
`interim-format.md`. Confirm the choice with the user.

### 2. Completeness gate

Run the completeness gate from `interim-format.md`. Refuse to render
from an incomplete package; list what is missing and stop.

### 3. Draft under the tier-3 contract

Read the deferred references above, then draft the case study in
this order — sanitized from the first word,
not sanitized afterwards:

1. **Context** — the kind of team and the kind of task, described
   without identifying anyone or anything internal.
2. **Approach** — what was built, in transferable terms.
3. **Results** — aggregates and ranges only; every figure tagged as
   tag plus method phrase per `audiences.md`; denominators and
   pinned-window lengths stated; honest rounding.
4. **Limitations** — candid and mandatory: measurement weaknesses,
   the counterfactual rung, what was not measured. Credibility is
   the PR asset.

### 4. Triangulate every headline claim

Apply the triangulation procedure from `evidence.md` to each
headline claim. A claim outside the defensible external range is
removed or kept only with an explicit methodology defense. Record
each outcome.

### 5. Final checklist pass

Run the tier-3 final checklist from `audiences.md` over the finished
draft as an explicit pass, item by item. Fix every hit, then write
`<run>/reports/case-study-public.md`.

## Output

Open with a one-line hero (`✓ Public case study: <headline with
tag>`, `✓ Evidence report: <supported claim, unknowns named>`, or
`⚠ Refused: <reason>`), then exactly these sections:

1. `## Run` — the run chosen and its pinned window.
2. `## Gate` — the completeness result.
3. `## Sanitization` — the checklist results item by item, and the
   triangulation outcome per headline claim.
4. `## Report` — where it was written and the headline.

End with an `ask-user-choice` panel: derive the PR release next,
tighten a claim, or stop. Skip the panel in plan mode or when
running non-interactively.


## Portability notes

- `ask-user-choice` — present the listed options and wait for the user to pick one. Hosts with a structured multiple-choice tool (Claude Code's `AskUserQuestion`) should use it; otherwise print a numbered list and wait for a numbered reply. Never proceed on an assumed answer.
- `$ARGUMENTS` — the text the user passed when invoking this skill. If your host does not substitute it, read it as the user's request in the current turn, and ask when there is none.
- Bundled files — every relative path in this skill points at a file shipped inside this skill directory. Read them from here, not from the host's plugin tree.