shared skills need a rollout plan too. An edit that helps one repo can quietly change how ten teams review code. I'd version the skill, test the new version against a small set of past failures, then let one team try it before changing the organization default. An owner and an update process are much more useful when there's a way back
The part about people hiding their AI tricks so the boss won't think they're cutting corners surprised me. I started Loomhouse in mid-October on the opposite habit, where every build goes out with the prompts and what it cost, and it hadn't occurred to me that inside a company the same openness could feel like a risk. Writing in public, the worst that happens is nobody reads it.
Project skills rot like READMEs do. A skill that still says `make test-fast` after the target got renamed leaves the agent guessing a substitute, and nobody sees an error. Treating skills like code should include a CI job that runs the commands each SKILL.md references.
Stack-ranking people on AI proficiency pays them to keep their tricks private, so celebrating shared workflows can only go so far. I think the fix belongs in the ranking itself. Count how many teammates use a skill you wrote, since skills already have owners. Then sharing raises your score instead of lowering it.
shared skills need a rollout plan too. An edit that helps one repo can quietly change how ten teams review code. I'd version the skill, test the new version against a small set of past failures, then let one team try it before changing the organization default. An owner and an update process are much more useful when there's a way back
That's the same reason I liked the staged-release and rollback layers in the Claude Code review setup I covered here: https://www.the-ai-corner.com/p/ai-code-review-claude-code-team
The part about people hiding their AI tricks so the boss won't think they're cutting corners surprised me. I started Loomhouse in mid-October on the opposite habit, where every build goes out with the prompts and what it cost, and it hadn't occurred to me that inside a company the same openness could feel like a risk. Writing in public, the worst that happens is nobody reads it.
Project skills rot like READMEs do. A skill that still says `make test-fast` after the target got renamed leaves the agent guessing a substitute, and nobody sees an error. Treating skills like code should include a CI job that runs the commands each SKILL.md references.
Stack-ranking people on AI proficiency pays them to keep their tricks private, so celebrating shared workflows can only go so far. I think the fix belongs in the ranking itself. Count how many teammates use a skill you wrote, since skills already have owners. Then sharing raises your score instead of lowering it.