# Case reports: how a report on one real project is written so that an absent reader can trust it

A first-pass multivocal review of 42 sources: 34 formal and 8 grey. It asks how a participant writes a case with repository counts, and what those counts miss. It names its baseline.

- Formal: 7 read in full or in sections, 17 by abstract, 5 located only.
- Grey: 8 read in part, all at or above 10 of 20; 2 tier 1 or 2 calls, 1 tier 1 book, 3 tier 3 posts.

## Files

| File | Role |
|---|---|
| [frame](frame.md) | step 1: the problem, every term once, and the field's name against the repository's |
| [protocol](protocol.md) | strings, stopping rule, selection and what was not searched |
| [sources](sources.md) | the flow, then each work with its grade or score |
| [review](review.md) | step 2: findings by strand |
| [state of the art](state-of-the-art.md) | step 3: the best evidence and the practice most adopted, per question |
| [against](against.md) | limits, controversies and the review's own weak points |
| [position](position.md) | step 4: what is known, what is not, what this repository adds |
| [baseline](baseline.md) | step 5: the guideline a case must meet |
| [bibliography](../bibliography.md) | every work cited, once, with a stable link |

## Claims

| Claim | Grade | Where |
|---|---|---|
| A report by those who took part is an experience report, not a case study in the sense of the community standard | convention | frame.md |
| The reader's checklist of Runeson 2009 has 12 items; the standard for case studies has 7 essential attributes | convention | review.md A |
| Version-control counts mislead: a repository is not a project, commits move and are rewritten, large commits differ in intent | measured, descriptive, on other datasets | review.md B |
| In the first case, 434 commits summed over seven repositories are 415 distinct by author time and subject | measured, in one case | against.md |
| Studies of coding agents measure a task, a pull request or a project; none in this pool counts questions that reached a person | measured, for each study; the gap is of this search | review.md D |
| Company write-ups of agent work report sessions and cost and not interventions | owner's claim | review.md D |
| No test shows that a reporting checklist makes a case more trusted | convention, untested | state-of-the-art.md Q2 |

## Named baseline

- **The reader's checklist of Runeson 2009**, items 39 to 50, read with the experience-report criteria of ICSE 2003 and the Case Study and Repository Mining standards.
- A case meets each item or says why not; a hand count is marked, and a program takes the rest ([baseline](baseline.md)).

## Status

A first pass: the five steps are written and the baseline is named. One reviewer; forward citation of three seed works was run (protocol section 11); the leads listed in [protocol](protocol.md) are not read.