Find code by asking what it does. A CLI for coding agents that uses Jev to discover relevant files and source context.
TypeScript ai-sdkclaude-codeclicode-searchcodexcoding-agentscontext-retrievaldeveloper-toolsjevsemantic-searchtypescriptvercel-ai-gateway
Run everything once, when a plan's implementation is finished; no milestone gates
David Zhang4d28fa1a85f6
diff --git a/.agents/skills/audit-agents/SKILL.md b/.agents/skills/audit-agents/SKILL.mdindex 57523af..77348f2 100644--- a/.agents/skills/audit-agents/SKILL.md+++ b/.agents/skills/audit-agents/SKILL.md@@ -25,7 +25,7 @@ AGENTS.md is read by everyone who will ever work on the project. Picture the nex- **A thousand people.** A newcomer can act on it without knowing this session, this author or this month's work.- **Changes what they do.** Delete a line a capable contributor already follows.- **Plain and concrete.** A principle is not an abstraction. "Look at the picture" works; "match verification to the claim" doesn't. Prefer a short sentence with its consequence, and one example from the product over a general noun.-- **Iteration speed.** This is the core principle and must be stated outright: optimize the time to feedback you can trust, run the narrowest check that answers the question, and keep the full gates for milestones the plan names in advance. Flag any rule that adds process to every loop, or puts the full gates before each commit, merge or push. Never trade away a real acceptance requirement for speed.+- **Iteration speed.** This is the core principle and must be stated outright: optimize the time to feedback you can trust, run the narrowest check that answers the question, and run everything once, when the plan's implementation is finished. Flag any rule that adds process to every loop, puts the full run before a commit, merge, push or finished feature, or names checkpoints for it along the way. The full run comes earlier only when the next piece of work can't be trusted without it. Never trade away a real acceptance requirement for speed.## Shapediff --git a/.agents/skills/audit-agents/assets/example-agents.md b/.agents/skills/audit-agents/assets/example-agents.mdindex 6e5cbf8..b2211c2 100644--- a/.agents/skills/audit-agents/assets/example-agents.md+++ b/.agents/skills/audit-agents/assets/example-agents.md@@ -18,9 +18,9 @@ Optimize for iteration speed. The measure is the time to feedback you can trust,Run the narrowest check that answers your question: one test, then one file, then one package. That is the proof for everyday work, including a commit, a merge and a push.-**The full gates are for milestones only.** Running everything is slow and saturates the machine, so it happens at a milestone the plan names in advance (a spec's stated checkpoint, a release) and once when a spec is closed. It is not a step before each commit, merge or push, and never a feedback loop. An agent working on one piece of a plan does not run it; whoever integrates the plan does, at the milestone.+**Run everything once, when a plan's implementation is finished.** Running everything is slow and saturates the machine. Until then a change is checked by what it can move: its own tests and the output it touches. That is enough for a commit, a merge, a push and a finished feature. A failure that only the full run finds is fixed at the end; that is cheaper than gating every step.-Between milestones, a change is checked by what it can move: its own tests and the output it touches. A failure found later at a milestone is fixed then; that is cheaper than gating every step.+A change that reaches the whole system does not bring that run forward on its own. While more work is coming, the full run still waits for the end. Run it sooner only when the next piece of work can't be trusted without it.Every expensive run must answer a question a cheaper one can't. <Name this product's expensive runs>; do only the ones a change can move. Reuse a result that is still valid, and rerun only what a change could have invalidated. Docs and data that no code reads need no run at all.diff --git a/AGENTS.md b/AGENTS.mdindex 0ff86f5..d3e5765 100644--- a/AGENTS.md+++ b/AGENTS.md@@ -18,9 +18,9 @@ Optimize for iteration speed. The measure is the time to feedback you can trust,Run the narrowest check that answers your question: one test, then one file, then one package. That is the proof for everyday work, including a commit, a merge and a push.-**The full gates are for milestones only.** Running everything is slow and saturates the machine, so it happens at a milestone the plan names in advance (a spec's stated checkpoint, a release) and once when a spec is closed. It is not a step before each commit, merge or push, and never a feedback loop. An agent working on one piece of a plan does not run it; whoever integrates the plan does, at the milestone.+**Run everything once, when a plan's implementation is finished.** Running everything is slow and saturates the machine. Until then a change is checked by what it can move: its own tests and the output it touches. That is enough for a commit, a merge, a push and a finished feature. A failure that only the full run finds is fixed at the end; that is cheaper than gating every step.-Between milestones, a change is checked by what it can move: its own tests and the output it touches. A failure found later at a milestone is fixed then; that is cheaper than gating every step.+A change that reaches the whole system (the retrieval pipeline, the response a command prints) does not bring that run forward on its own. While more work is coming, the full run still waits for the end. Run it sooner only when the next piece of work can't be trusted without it.Every expensive run must answer a question a cheaper one can't. Benchmark runs, which spend real money, and the container tests of the installed package are the expensive runs here; do only the ones a change can move. Reuse a result that is still valid, and rerun only what a change could have invalidated. Docs and data that no code reads need no run at all.