From 6d9e8a162aa55674dd8168df41e218d9875809ca Mon Sep 17 00:00:00 2001 From: Fleet Sync Bot Date: Wed, 23 Sep 2026 12:46:29 -0700 Subject: [PATCH] docs: sync fleet-managed agent sections from Repository_Management Propagates the fleet-managed sections from the central Repository_Management/AGENTS.md into both agent rule surfaces: AGENTS.md (Codex and the universal fallback) and CLAUDE.md (Claude Code). Covers Reasoning & Engagement, Network & API Hygiene, Codemap Freshness, Durable Implementation Handoffs, Development Logs, and Agent Presence and Communication. --- AGENTS.md | 15 +++++++++++---- CLAUDE.md | 15 +++++++++++---- 2 files changed, 22 insertions(+), 8 deletions(-) diff --git a/AGENTS.md b/AGENTS.md index 27702bf..3499666 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -445,10 +445,12 @@ never as a silent close, never as a fabricated result. ### What to Do When You Hit One -1. **Do not attempt the measurement and do not simulate it.** Never write a - number into an issue, a test, a report or a commit that was not measured. -2. **Do not close the issue.** Propose a record in the repository's - `docs/planning/deferred-validation.json`: the original issue URL, its +1. **Never fabricate a measurement.** Explicitly synthetic numerical fixtures + remain useful software tests; they cannot satisfy physical acceptance. +2. **Do not close first.** Read `docs/development/planning/catalog.json` for + an existing v1 adopter; otherwise use `docs/planning/deferred-validation.json`. + Maintain one authority; never create a competing empty catalog. Preserve the + original issue URL, its acceptance criteria copied verbatim, why it is external, the resources it needs, and what would reopen it. Validate with `python -m shared_scripts.deferred_validation --repo-root .`. @@ -459,6 +461,11 @@ never as a silent close, never as a fabricated result. on the default branch, and a named human or the Board signed it off (`reviewed_by` / `reviewed_on`). +Existing deployed plans retain their original source snapshots, pending Board +resource decisions and reactivation requirements. A schema change supplies no +review approval or measurement. Do not reopen already deferred `roadmap` work +because its experiment remains undone; inspect the plan and publication receipt. + ### Discovery Is Not Authority Matching keywords like `measure`, `rig`, `chamber`, `specimen`, `calibrat` or diff --git a/CLAUDE.md b/CLAUDE.md index f23841f..0d4b337 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -445,10 +445,12 @@ never as a silent close, never as a fabricated result. ### What to Do When You Hit One -1. **Do not attempt the measurement and do not simulate it.** Never write a - number into an issue, a test, a report or a commit that was not measured. -2. **Do not close the issue.** Propose a record in the repository's - `docs/planning/deferred-validation.json`: the original issue URL, its +1. **Never fabricate a measurement.** Explicitly synthetic numerical fixtures + remain useful software tests; they cannot satisfy physical acceptance. +2. **Do not close first.** Read `docs/development/planning/catalog.json` for + an existing v1 adopter; otherwise use `docs/planning/deferred-validation.json`. + Maintain one authority; never create a competing empty catalog. Preserve the + original issue URL, its acceptance criteria copied verbatim, why it is external, the resources it needs, and what would reopen it. Validate with `python -m shared_scripts.deferred_validation --repo-root .`. @@ -459,6 +461,11 @@ never as a silent close, never as a fabricated result. on the default branch, and a named human or the Board signed it off (`reviewed_by` / `reviewed_on`). +Existing deployed plans retain their original source snapshots, pending Board +resource decisions and reactivation requirements. A schema change supplies no +review approval or measurement. Do not reopen already deferred `roadmap` work +because its experiment remains undone; inspect the plan and publication receipt. + ### Discovery Is Not Authority Matching keywords like `measure`, `rig`, `chamber`, `specimen`, `calibrat` or