| name | case-study-writer |
| description | Write public-safe case studies from private projects, automations, agent workflows, internal tools, demos, or client-facing systems. Use when the user asks to turn project work into a case study, create client proof, write a sanitized GitHub case study, summarize an internal system publicly, document an automation without sharing source code, or explain a workflow while removing credentials, raw prompts, logs, databases, client data, local paths, and private implementation details. |
Case Study Writer
Overview
Use this skill to turn private project work into public-safe proof: concise case studies that explain the problem, workflow, outcome, and adaptation value without exposing source code, credentials, raw prompts, logs, databases, client data, or private implementation details.
Workflow
- Clarify the public artifact target: GitHub case study, website system page, portfolio note, demo companion, or README section.
- Read
references/case-study-template.md and use its section order unless the destination already has a stronger house style.
- Read
references/safety-boundary-examples.md before writing the opening boundary note and final Safety Boundary section.
- Extract only public-safe facts:
- Problem the system solves.
- What the system does at workflow level.
- Human review and approval points.
- General stack categories when useful.
- Outcome stated qualitatively or with approved aggregate numbers.
- What a similar client could adapt.
- Rewrite private details into generic, reusable language. Do not copy raw notes, prompts, logs, exports, database rows, or source code.
- Produce the case study in Markdown and include a short safety verdict:
BLOCK, REVIEW, or READY.
Case Study Shape
Use these sections by default:
- Title
- Boundary note
- Public page or demo links, if already approved and live
- Problem
- What The System Does
- Workflow
- Outcome
- What Can Be Adapted For Clients
- Safety Boundary
Keep the case study short enough to scan quickly. Prefer six to eight workflow steps and three to five client-adaptation bullets.
Safety Rules
- Never include source code, private prompts, API keys, credentials, databases, logs, exports, client messages, account identifiers, local paths, or screenshots with private UI.
- Do not name clients, prospects, internal codenames, domains, or people unless the user confirms they are approved for public use.
- Use broad stack categories such as Python, SQLite, browser automation, queue dashboard, or LLM API instead of exact private architecture when details are not needed.
- Mention human approval explicitly when the system drafts, scores, publishes, sends, spends, or changes a system of record.
- If the user provides risky source material, return
BLOCK with removal instructions before writing the publishable draft.
Output Format
Return:
- Safety verdict:
BLOCK, REVIEW, or READY.
- Short note on what was excluded or generalized.
- The Markdown case study.
- Remaining checks before publication.
If the user asks you to edit files, write the draft into the requested location or a sanitized staging folder, then recommend running a public-safety scan before commit.