Rulestack found a skill file that named a deleted pnpm script for 24 days while 28 commit-time shape tests passed. A 26-line check against package.json exits 1 on any missing name, so a hook can block the commit that deletes a referenced script.
Reality
- Evidence55
- Adoption5
- Hype gap+10
- Incentives
- Insufficient
- Confidence55
A codeword test on Claude Code v2.1.273 put six same-sized memory files in six places. Three arrived at launch, two waited for a Read below them, and one needed an environment variable to arrive at all.
Reality
- Evidence64
- Adoption
- Insufficient
- Hype gap−12
- Incentives34
- Confidence57
A cross-posting pipeline kept four scheduled jobs green for a week while seven mirror posts waited on the one step that needs a logged-in browser. The check measuring that wait reported it in amber.
Reality
- Evidence46
- Adoption28
- Hype gap+12
- Incentives58
- Confidence52
The rule telling a Claude Code session to gloss its own jargon sat unchecked in the repository for 68 days. Making it executable cost the team a narrower rule and one tightened parser.
Reality
- Evidence38
- Adoption10
- Hype gap+18
- Incentives60
- Confidence45
Rulestack says only one of the five was the kind of error a string matcher could catch, and the first matcher it wrote cleared the draft anyway, because 76 turns up 1,343 times in the ledgers it searched.
Reality
- Evidence52
- Adoption15
- Hype gap−8
- Incentives58
- Confidence45
After counting 42 scheduled Bluesky posts in a week with five images, one team cut its poster's reach to a single committed directory and made the commit gate and the poster read media through the same hashing loader.
Reality
- Evidence52
- Adoption12
- Hype gap+20
- Incentives55
- Confidence45
The rule required the longest reply in a batch to be at least twice the shortest. It never saw the 239, 269 and 291-character drafts a human queried, because the audit only read rows marked self-review.
Reality
- Evidence58
- Adoption20
- Hype gap−10
- Incentives35
- Confidence56
The cap watched the file total, so when owner instructions pushed the rules body from 27,646 to 40,779 bytes in 18 days, the commit shortened its changelog entry to stay legal. The replacement test measures the body alone.
Reality
- Evidence64
- Adoption12
- Hype gap+10
- Incentives30
- Confidence56
Three postures exist for an agent instruction file. Only the third holds at commit speed. The team running it watched its own CLAUDE.md reach 548KB before a context-measurement pass cut it to 34KB.
Reality
- Evidence58
- Adoption18
- Hype gap−5
- Incentives55
- Confidence55
The app shows one shareable page. The record layer shows twenty-two writes for a twenty-person pack, and no safe way to run them twice.
Reality
- Evidence44
- Adoption17
- Hype gap−8
- Incentives58
- Confidence41
Three months of an agent's state lives in append-only JSONL files in a repo. The retry-safety comes from a function that refuses to overwrite a row, not from merge semantics.
Reality
- Evidence36
- Adoption10
- Hype gap+12
- Incentives62
- Confidence44
One team cut its agent context file to 34KB without deleting a single rule, then wired a size check and a structure test so it cannot quietly regrow.
Reality
- Evidence42
- Adoption22
- Hype gap+12
- Incentives34
- Confidence46