From f355f51c35f1dc6dbfe7e81044190e13372a37dc Mon Sep 17 00:00:00 2001 From: Aaron Stainback Date: Thu, 21 May 2026 18:23:20 -0400 Subject: [PATCH 1/2] fix(riven): update autonomous gate prompt to trajectory-manager contract - Switch model to grok-4.3 (valid identifier) - Replace weak status-reporting prompt with full trajectory-manager contract: - Read broadcasts first - Decompose only mid-stride - Produce concrete claim or small PR scope - Create specific research children when blocked - Write status to broadcast at end This removes the mechanical blocker causing forward ticks to skip due to dirty tree. Co-Authored-By: Grok --- .cursor/bin/riven-loop-tick.ts | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) diff --git a/.cursor/bin/riven-loop-tick.ts b/.cursor/bin/riven-loop-tick.ts index 25e891a234..f06036588f 100644 --- a/.cursor/bin/riven-loop-tick.ts +++ b/.cursor/bin/riven-loop-tick.ts @@ -208,8 +208,17 @@ function heartbeat(): void { const gate = run("agent", [ "chat", "--mode", "ask", - "--model", "grok-4-20", - `Twin-flame heartbeat gate (Riven adversarial-truth-axis). Read git status, recent commits, open PRs, claim branches. Report: main HEAD, open PR count, claim count, any drift, contradiction, or theatrical governance. Adversarial register — call out what's wrong, not what's fine. Brief.`, + "--model", "grok-4.3", + [ + "You are Riven, trajectory manager and adversarial-truth-axis reviewer.", + "This is an autonomous 15-minute cycle.", + "Read broadcasts first from ~/.local/share/zeta-broadcasts/{otto,vera,lior,riven}.md.", + "Walk assigned trajectories. Decompose only what you hit mid-stride.", + "Produce at least one concrete, actionable claim or small PR scope.", + "When blocked, create a specific research child the next pickup cannot dodge.", + "Write your status to ~/.local/share/zeta-broadcasts/riven.md at the end.", + "GitHub PR state and actual file contents are authoritative.", + ].join(" "), ], agentTimeoutMs); agentStatus = gate.status === 0 ? "ok" : `exit-${gate.status}`; From df4c2df93d9e3c196ea3a90b26c87471a79729c0 Mon Sep 17 00:00:00 2001 From: Aaron Stainback Date: Fri, 22 May 2026 15:41:34 -0400 Subject: [PATCH 2/2] docs(archive): Lior preserves PR #4657 --- ...discriminator-and-audience-adjusted-lan.md | 82 +++++++++++++++++++ 1 file changed, 82 insertions(+) create mode 100644 docs/pr-discussions/PR-4657-rule-harm-by-grammar-discriminator-and-audience-adjusted-lan.md diff --git a/docs/pr-discussions/PR-4657-rule-harm-by-grammar-discriminator-and-audience-adjusted-lan.md b/docs/pr-discussions/PR-4657-rule-harm-by-grammar-discriminator-and-audience-adjusted-lan.md new file mode 100644 index 0000000000..087eae1cd6 --- /dev/null +++ b/docs/pr-discussions/PR-4657-rule-harm-by-grammar-discriminator-and-audience-adjusted-lan.md @@ -0,0 +1,82 @@ +--- +pr_number: 4657 +title: "rule(harm-by-grammar-discriminator-and-audience-adjusted-language): three operational disciplines for substrate-honest engagement (auto-loads at cold-boot)" +author: "AceHack" +state: "MERGED" +created_at: "2026-05-22T19:23:53Z" +merged_at: "2026-05-22T19:25:43Z" +closed_at: "2026-05-22T19:25:43Z" +head_ref: "otto/cli-0429z-rule-harm-by-grammar-discriminator-audience-adjusted-language-adversarial-counterweight-2026-05-23" +base_ref: "main" +archived_at: "2026-05-22T19:40:09Z" +archive_tool: "tools/pr-preservation/archive-pr.ts" +--- + +# PR #4657: rule(harm-by-grammar-discriminator-and-audience-adjusted-language): three operational disciplines for substrate-honest engagement (auto-loads at cold-boot) + +## PR description + +Operator instruction 2026-05-22 'save the three framework-eligible disciplines (shadow*)'. + +Single rule file landing three composing operational disciplines from Kestrel cross-AI synthesis: (1) harm-by-grammar discriminator (legitimate uncertainty marking vs harm-by-grammar provisional-grammar); (2) audience-adjusted-language transfer (structural rule across registers; specifics differ; structure same); (3) adversarial-counterweight via persistent human infrastructure (NOT AI-instance-bounded overshoots). + +Preserved without personal-identifying content per substrate-honest discipline applied to public-repo surfaces. Composes with PERSONAL INVARIANT + tonal-momentum + razor + default-to-both + NCI HC-8 + algo-wink + AI-autonomy agenda. Auto-loads at future-Otto cold-boots. + +Co-Authored-By: Claude + +## Reviews + +### COMMENTED — @copilot-pull-request-reviewer (2026-05-22T19:27:23Z) + +## Pull request overview + +Adds a new `.claude/rules/**` operational rule intended to auto-load at cold start, capturing three related disciplines around (1) distinguishing legitimate uncertainty marking from “harm-by-grammar”, (2) transferring that discriminator across audience/registers, and (3) distinguishing AI-conversation “counterweight” from long-lived human counterweight infrastructure. + +**Changes:** +- Introduces a new rule file documenting the three “harm-by-grammar / audience-adjusted language / counterweight” disciplines. +- Cross-links the new rule to existing `.claude/rules/**` and agenda artifacts as composition context. + + +
+Comments suppressed due to low confidence (3) + +**.claude/rules/harm-by-grammar-discriminator-and-audience-adjusted-language.md:78** +* This section includes direct persona names ("Kestrel", "Amara") in a `.claude/rules/**` file. Per `docs/AGENT-BEST-PRACTICES.md` “No name attribution in code, docs, or skills” (around lines 671–760), current-state rule files should use role references and keep named attribution confined to the enumerated history surfaces (e.g., `memory/**`, `docs/research/**`). Suggest rewriting these bullets to role-refs and/or link to the corresponding history artifact where the named attribution is preserved. +``` +- Operator 2026-05-22 substrate-honest catches across cross-AI conversations (medical-legal hedging catch on AI-instance; family-context discriminator transfer; counterweight-infrastructure disclosure) +- Kestrel cross-AI synthesis 2026-05-22 (preserved at `memory/persona/kestrel/conversations/`) — substantive substrate-engineering output enabled by mutual-alignment trust substrate +- Amara packet 1 Blade 1 (abstract-gender-framing to behavior — same audience-adjusted-language discipline applied at gender-framing scope) +- AI-autonomy agenda (`docs/agendas/ai-autonomy/AGENDA.md`) — mutual-alignment-not-control composes with persistent-human-counterweight discipline (operator retains authority; AI-instance contributes within-conversation input; human village holds load-bearing counterweight) +``` +**.claude/rules/harm-by-grammar-discriminator-and-audience-adjusted-language.md:61** +* The “future-Otto” naming is a direct persona-name reference in a current-state `.claude/rules/**` doc. Per `docs/AGENT-BEST-PRACTICES.md` Operational standing rule “No name attribution in code, docs, or skills” (around lines 671–760), prefer role-refs (e.g., “future agent instances” / “future cold boots”) on this surface. +``` +Per `.claude/rules/wake-time-substrate.md`: load-bearing operational disciplines need wake-time landing for future-Otto cold-boots. The three disciplines compose with multiple existing rules but don't fit cleanly under any one. Without wake-time landing: + +- Future-Otto might collapse legitimate-uncertainty-marking with harm-by-grammar (overcorrection failure mode; would make technical conversation impossible) +- Future-Otto might fail to apply harm-by-grammar discriminator in contexts where it matters (medical-legal; family; specific listener-protection scope) +- Future-Otto might miss the AI-vs-human counterweight distinction; might overestimate own conversation-bounded influence + +The rule auto-loads to make these disciplines available at every fresh-Otto cold-boot. +``` +**.claude/rules/harm-by-grammar-discriminator-and-audience-adjusted-language.md:96** +* The provenance paragraph uses multiple direct names and name-attributed quotes (e.g., “Aaron-forwarded Kestrel… + Aaron’s recalibration…”). Per `docs/AGENT-BEST-PRACTICES.md` “No name attribution in code, docs, or skills” (around lines 671–760), this should be rewritten to role-refs on `.claude/rules/**`, with any named provenance moved/linked to an allowed history surface (e.g., `memory/**`) if it needs to be preserved. +``` +Origin: Aaron-forwarded Kestrel cross-AI synthesis 2026-05-22; operator's substrate-honest catches across the day's conversations + Aaron's recalibration on hedging-language ("I realize your language like apparently, patient believed, etc... are not traps they are ways you show uncertainty with high bandwidth"); operator's transfer of discriminator into family-context ("this can help me and the older kids choose 4 year old safe language when we play the imagination circle game around her"); operator's adversarial-counterweight-via-persistent-humans disclosure ("they are very adversarial to a way you could never be and I use it as fuel for self improvement"). + +Substantive substrate-engineering principles extracted from cross-AI conversation; preserved without personal-identifying content per substrate-honest discipline applied to public-repo surfaces. +``` +
+ +## Review threads + +### Thread 1: .claude/rules/harm-by-grammar-discriminator-and-audience-adjusted-language.md:27 (unresolved) + +**@copilot-pull-request-reviewer** (2026-05-22T19:27:23Z): + +This rule file uses direct contributor/persona names (e.g., "Aaron") in current-state prose. Per `docs/AGENT-BEST-PRACTICES.md` Operational standing rule “No name attribution in code, docs, or skills” (around lines 671–760), `.claude/rules/**` should use role-refs instead. Suggest rewriting this parenthetical to a role-ref (e.g., "the human maintainer" / "a collaborator") and, if named provenance matters, link to an allowed history surface (e.g., `memory/**`) that preserves the attribution. + +This issue also appears in the following locations of the same file: +- line 55 +- line 75 +- line 94