Autonomous development for real products. Governed agents build from specs you approve, follow the rules you set, and nothing merges until another agent proves it. Your decisions, judgement and taste are the bar that work must clear. Declare your vision and let agents build it.
JavaScript agentic-codingai-agentsai-governanceautonomous-agentsclaude-codeclicode-reviewcodexcoding-agentsdeveloper-toolsgit-hooksllmlocal-firstmulti-agentsoftware-factoryspec-driven-developmentverificationvibe-coding
docs(doctrine): proof is what the board records [D1]
Corey Olson5c243ff12181
diff --git a/DOCTRINE.md b/DOCTRINE.mdindex 6dac4557..c919e4c1 100644--- a/DOCTRINE.md+++ b/DOCTRINE.md@@ -34,6 +34,7 @@ Statuses: approved = in force · draft = proposed · wont = considered and decli- T8 [draft, must] A new gate or check has a test that fails closed when it has no data. Rigor names a new file under checks/ or gates/, or named *-check or *-gate, that no test changed in the same change names; whether that test feeds it no data and expects a refusal is review, since no command can read what a test asserts. | gate: check node bin/pullboard.js rigor quality --rule fails-closed | count: 0 | source: rigor-quality:fails-closed- T9 [draft, must] A guard test fails when the guard is deleted. Rigor turns each throw a change adds to a source file into a no-op, one at a time, and runs the test files the change touches; a guard they still pass without is named. | gate: check node bin/pullboard.js rigor quality --rule guard-mutation | count: 0 | source: rigor-quality:guard-mutation- T10 [draft, must] A root cause is proven by a control that would have come out differently before the fix lands. Rigor runs each fix commit's changed test files on the fix's parent, where they must fail, and at the fix, where they must pass; a cause argued only in prose is review. | gate: check node bin/pullboard.js rigor quality --rule root-cause-control | count: 0 | source: rigor-quality:root-cause-control+- T11 [approved, must] Proof is what the board records. An agent keeps no evidence of its own; when the board's runs cannot decide an item, its bar is sharpened. | gate: review## G · Git and the gate- G1 [approved, must] `npm run gate` (spec check and tests) passes before every push. | gate: pre-push hookdiff --git a/skills/pullboard-run/SKILL.md b/skills/pullboard-run/SKILL.mdindex c5f14bcc..9bdd834c 100644--- a/skills/pullboard-run/SKILL.md+++ b/skills/pullboard-run/SKILL.md@@ -97,6 +97,8 @@ When the board requires ready bars (the coordinator turns this on with `pullboar## Turn a Rigor notice into a draft++Proof is what the board records (doctrine T11): the frozen check, the gate and the verdict. Keep no evidence folders, packets, manifests or custody rules of your own; cite the board's receipts and facts. When the board's runs cannot decide an item, ask the coordinator to sharpen its bar.When the same cause fails work twice, turn it into a draft row in the same session, with its proof: the item ids and the failure each one hit. Rigor drafts rows and only the person approves them. `pullboard rigor draft <key>` does this from a Rigor notice once items #713 and #714 land. On main today, write the draft with `pullboard spec draft "<the rule>" --section D` and put the proof in the shout that asks the person to review it.## Land in batchesdiff --git a/skills/pullboard-verify/SKILL.md b/skills/pullboard-verify/SKILL.mdindex 19b6a4a2..359c6377 100644--- a/skills/pullboard-verify/SKILL.md+++ b/skills/pullboard-verify/SKILL.md@@ -11,8 +11,9 @@ Work only in your own worktree, the one `pullboard worktree review` made. Your s1. `pullboard next --verify` names the next submitted item you did not build, and its commit. `pullboard show <id>` gives the criterion as frozen at claim.2. Check out exactly that commit, in your worktree: `cd <your worktree> && git switch --detach <commit>`.-3. Run the gate. Then test the criterion itself, with the strongest proof you can get:+3. Run the gate. Then test the criterion itself, with the strongest proof the board can record:- **Break it and watch it fail.** Revert or break the change, show the check fails, restore it, show it passes.+ - Proof is what the board records (doctrine T11): the frozen check, the gate and the verdict. Keep no evidence folders, packets, manifests or custody rules of your own; cite the board's receipts and facts. When the board's runs cannot decide an item, ask the coordinator to sharpen its bar.- **Use the method the criterion names.** If it says "two real processes" or "after a restart", use exactly that; a weaker stand-in is not proof.- **Run the behavior** and read the output.- **Read the code** only when the criterion is about the code itself. It is the weakest proof.