doctrine(authority): standing authority — create public test git repos, full admin, hourly billing tracking

memory/MEMORY.md

@@ -2,6 +2,7 @@
 
 **📌 Fast path: read `CURRENT-aaron.md` and `CURRENT-amara.md` first.** <!-- latest-paired-edit: fork-audit R/C/T diff-filter coverage + plumbing-vs-porcelain note (2026-04-29 round-10 Amara). NOTE: this comment is a single-slot "latest paired edit" marker (not a paired-edit log). Per the round-10 Amara framing the slot semantics are now explicit. -->
 
+- [**Standing authority — create public test git repos on AceHack + LFG, full admin, hourly billing tracking (Aaron, 2026-04-29)**](feedback_standing_authority_create_test_git_repos_public_only_track_billing_aaron_2026_04_29.md) — *"you have standing authority at any time to create git repos on acehack and lfg to test any features of git they just have to be public cause that's free... full admin... just track the billing every hour"* + clarification *"not noticing and stopping costs until we talk is the barrier, a mistaken accident spend is fine if you are auditing billing and catch the costs that way."* Standing grant: agent creates test repos on either org at any time (no per-creation Aaron sign-off), full admin to exercise any git/GitHub/CI/Actions/branch-protection/ruleset feature, with TWO binding constraints — public only (free; never private), and hourly billing tracking must cover the new repos (audit-and-catch is the safety mechanism, not pre-perfect cost-zero). Failure mode is **silent spend**, not spend itself: audit-coverage is more load-bearing than spend-zero. Composes with Otto-365 "basically never ask" (test-repo creation IS invariant maintenance), branch-protection-settings-are-agent-call (delegated authority pattern), task #315 (hourly budget cadence — load-bearing safety latch), task #287 (cost visibility), AceHack mirror-not-peer doctrine (mirror constraint applies to AceHack/Zeta specifically; AceHack as ORG can host test repos), Aaron's visibility-constraint rule (test repos are inherently visible + billing surface = both legs hold).
 - [**Otto-364 — Search-first for authoritative claims, not training data, not project memory (Aaron, 2026-04-29)**](feedback_otto_364_search_first_authority_not_training_data_not_project_memory_aaron_2026_04_29.md) — *"Training data is historical. Project state is historical. Current upstream docs are the test. Search first. Cite second. Assert third."* Generalises Otto-247 (version-currency) to ALL authoritative claims (tools / standards / APIs / runtimes / libraries / CI services / security policies). When asserting a load-bearing claim about anything upstream, WebSearch first, cite (URL + date searched), then assert. Project-state grep is a cross-check input, NOT a substitute. Demonstration via 4 web-search verifications of Amara's CI-classifier claims (Bun ci/lockfile, GitHub Actions paths-ignore + outputs, mise config) — each search produced a *sharper finding* than training-data recall. Verbatim packet + verifications at `docs/research/2026-04-29-aaron-search-first-authority-not-training-data-not-project-memory.md`. Composes with Otto-247 (version-specific predecessor — NOT superseded), Otto-363 (search results in chat = weather; cited in research doc = substrate), Otto-362 (stale claims must be refreshed — Otto-364 is upstream-vs-recall version), best-practices-evidence-lineage rule.
 - [**Otto-363 — Substrate or it didn't happen — no invisible directives (Aaron + Amara, 2026-04-29; refined by 5-AI review)**](feedback_otto_363_substrate_or_it_didnt_happen_no_invisible_directives_aaron_amara_2026_04_29.md) — *"A directive that lives only in a conversation is not a directive. It is weather. Substrate or it didn't happen. But also: indexed, reachable, and reconstructable — or it is not substrate yet. If you cannot point to the substrate, you are not done. You are just currently convinced."* Substrate is committed + reachable + indexed (all three legs). 5-tier channel taxonomy: ephemeral (chat/TaskUpdate/`/tmp`/`/var/tmp` — NEVER call done) / local-parked (named stash, local WIP) / remote-parked (pushed WIP branch, draft PR — *"if it matters enough to come back to, it deserves a git ref"*) / host-durable-not-git-canonical (GitHub Issues, PR comments) / git-native-preserved (committed + reachable-from-long-lived-ref + indexed repo files). 8-mechanism remediation: detector / verbatim-preservation paired with structured extraction / magnitude classifier (small/implementation/doctrine/superseding) / supersession protocol (bidirectional `supersedes:`/`superseded_by:` metadata, top-of-file stale banner OR quarantine to archive — NOT bottom-append; per Otto-362 generalisation) / cold-start proof (six questions including context-loss check) / "done"-vocabulary lock (captured ≠ parked ≠ preserved ≠ canonical ≠ operational, plus preserved-but-disputed) / CLAUDE.md+AGENTS.md bootstrap pointer / vocabulary-enforcement trailer (`Durability:`/`Substrate:`) eventually lintable. Default preservation route when uncertain: `docs/research/` first. Verbatim packets at `docs/research/2026-04-29-amara-substrate-or-it-didnt-happen-mechanisms-against-substrate-loss.md` (original) and `docs/research/2026-04-29-amara-substrate-or-it-didnt-happen-5ai-review-wave-corrections.md` (5-AI review wave + 10 review corrections; numbering matches the structured extraction). Composes with Otto-362 (intra-file supersession), channel-verbatim preservation, no-directives-otto-prose lint, verify-before-deferring/future-self-not-bound/never-be-idle/version-currency (all CLAUDE.md-tier), AND task #321 (git-recovery process — `wip/<topic>-<date>` parking branches are discoverable by name pattern; recovery process treats them as WIP-INTENTIONAL, not lost; complete parking + recovery loop is mechanical not vigilance-based).
 - [**Otto-362 — Doctrine memory expansion refreshes stale statements in the SAME edit (2026-04-29)**](feedback_otto_362_doctrine_memory_expansion_refresh_stale_statements_same_edit_2026_04_29.md) — When a memory file gets expanded with a new section that supersedes earlier statements in the same file, refresh the now-stale statements in the same edit, not a follow-up tick. Internal contradictions within one file are lying-by-omission. Pattern observed across 4 same-day doctrine PRs (#850/#851/#852/#853) where multi-AI review caught the drift instead of pre-push self-audit (10+ Copilot P1 + Codex P2 threads, all stale-statement class). Composes with same-tick CURRENT-update discipline (intra-file generalisation), verify-before-deferring, future-self-not-bound. Editing discipline, not lint — semantic contradictions can't be mechanised.