← ClaudeAtlas

rigorlisted

Set the rigor profile for atomic runs in this project — lean, standard, or thorough — scaling effort budgets (turns, loops, repairs) to task risk. Gate strength is not configurable.
uppifyagency/atomic-cc · ★ 0 · AI & Automation · score 63
Install: claude install-skill uppifyagency/atomic-cc
Set the atomic-cc rigor profile for this project. A profile scales EFFORT BUDGETS to the task's risk. It does not, and cannot, change how strong the review gate is. | Profile | max_turns / max_loops | verifier_count* | max_repairs | When | |---|---|---|---|---| | lean | 5 | 1 | 1 | Small low-risk changes, prototypes | | standard | 10 | 3 | 2 | Default engineering work | | thorough | 20 | 5 | 4 | Money paths, auth, data migrations | \* `verifier_count` applies only to `adversarial-verification`, whose upstream default is 3 (clamp 1–5). It does **not** apply to `goal` or `ralph`: goal's review quorum is 2 and ralph's reviewer count is 2, both module constants upstream and constants in this port. No profile can reduce either gate to a single reviewer — "unanimity 1/1" is not a gate, and a rigor setting that could weaken verification would be worse than no setting at all. How the profile actually reaches a run (this is the part that used to be missing): 1. `bin/rigor.sh set <profile>` writes `.atomic-cc/config.json`. 2. The plugin's SessionStart hook (`bin/rigor.sh notice`) prints the active profile and its budgets into session context, so the assistant invoking a workflow sees them without reading any file. 3. When you invoke an atomic workflow in this project, pass those budgets as workflow args. `bin/rigor.sh args` prints the exact JSON fragment to merge. Steps: 1. Parse $ARGUMENTS. With `lean`, `standard`, or `thorough`, run `"${CLAUDE_PLUGIN_ROOT}"/bin/rigor.sh set <pr