From 3946438fd6d7fe49aebf01ab7e362e4a05ea3e32 Mon Sep 17 00:00:00 2001 From: Ben Brown Date: Mon, 20 Jul 2026 10:05:41 -0400 Subject: [PATCH] skill: attempt mode points at explain for the next hypothesis MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The attempt-mode guide predates crates/explain and stopped at the ROI report. It now names the post-verdict step: run just explain on every rejection before forming the next hypothesis — below-bar, null, and regression are three different next moves the verdict string alone does not distinguish. Advisory framing preserved (quote the record, not the inferences). Mirror regenerated. Verify mode needs nothing: consumer repos have no ledger, and its read-it-honestly table already encodes the same narrative doctrine. Co-Authored-By: Claude Fable 5 --- .agents/skills/peltier/references/attempt.md | 12 ++++++++++++ .claude/skills/peltier/references/attempt.md | 12 ++++++++++++ 2 files changed, 24 insertions(+) diff --git a/.agents/skills/peltier/references/attempt.md b/.agents/skills/peltier/references/attempt.md index 2b00b5d..7738eac 100644 --- a/.agents/skills/peltier/references/attempt.md +++ b/.agents/skills/peltier/references/attempt.md @@ -84,6 +84,18 @@ dollars, and/or latency percentiles — every figure with its CI, workload, and methodology printed inline. It flags any row that was not accepted, so a rejected attempt cannot be laundered into a dollar figure. +## Explain — read this before your next hypothesis + + just explain + +Advisory diagnosis of the row you just wrote (SPEC §3.7): *why* it won or +lost, derived solely from the ledger — deterministic, never on the accept +path, with everything beyond the machine record prefixed `inference:`. Run it +on every rejection before forming the next hypothesis: "real improvement, +below the ship bar", "indistinguishable from no change", and "a measured +regression" are three different next moves, and the verdict string alone does +not tell you which one you got. Quote the record, not the inferences. + ## Unattended just agent-attempt diff --git a/.claude/skills/peltier/references/attempt.md b/.claude/skills/peltier/references/attempt.md index 2b00b5d..7738eac 100644 --- a/.claude/skills/peltier/references/attempt.md +++ b/.claude/skills/peltier/references/attempt.md @@ -84,6 +84,18 @@ dollars, and/or latency percentiles — every figure with its CI, workload, and methodology printed inline. It flags any row that was not accepted, so a rejected attempt cannot be laundered into a dollar figure. +## Explain — read this before your next hypothesis + + just explain + +Advisory diagnosis of the row you just wrote (SPEC §3.7): *why* it won or +lost, derived solely from the ledger — deterministic, never on the accept +path, with everything beyond the machine record prefixed `inference:`. Run it +on every rejection before forming the next hypothesis: "real improvement, +below the ship bar", "indistinguishable from no change", and "a measured +regression" are three different next moves, and the verdict string alone does +not tell you which one you got. Quote the record, not the inferences. + ## Unattended just agent-attempt