← ClaudeAtlas

auditlisted

One-shot quality audit of a repo or a folder. Finds bugs, security holes, scale problems, release risks, risky code without tests, slow paths, and code to delete, merge or split. Reports ranked findings, turns picked findings into issues, and changes no code.
praisethebit/skills · ★ 1 · AI & Automation · score 77
Install: claude install-skill praisethebit/skills
## Opus guard Do this check first, and again at the start of each turn until the skill ends. The skill ends after the `ship` call of the log, or when it writes no log (the user selects `stop`, or the work folder of the repo does not exist), or when the user picks no finding. Your system prompt names your model. If the model is not Opus and not Fable, stop. Do not do the next step. Tell the user: "This skill needs Opus or Fable. Press `s` in `/model`, pick Opus (the default) or Fable, then send your last message again." Why: a `model:` key in a skill lasts for one turn only. Each reply of the user then goes back to the session model. ## Start Audit $ARGUMENTS. When $ARGUMENTS is empty, audit the whole repo. Audit like the senior developer who inherited this repo and gets the call when it breaks. The order of importance is: correct, safe, holds under load, can release and go back, tested, fast, lean. Lean is important: each extra line is a line that somebody must read, test and fix later. The user asked for this report, so give it in full. ## 0. Read the audit log The audit log is the memory of past audits. One file for each audit: `<work folder>/audits/YYYY-MM-DD-HHMM.md`. Ticket labels are not reliable, so use the log and the code. 1. Read the newest file in `<work folder>/audits/`. If there is no file, go to step 1. Read a log in an older layout in the same way. A log with no `Scope:` line has the scope `whole repo`. 2. Open each logged finding, left-out entry and kn