neo-agent-skills 0.1.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.agents/skills/architecture-pre-flight/SKILL.md +10 -0
- package/.agents/skills/architecture-pre-flight/references/architecture-pre-flight-workflow.md +48 -0
- package/.agents/skills/blocked-task-state/SKILL.md +10 -0
- package/.agents/skills/blocked-task-state/references/blocked-task-state-workflow.md +64 -0
- package/.agents/skills/blog-post/SKILL.md +10 -0
- package/.agents/skills/blog-post/references/blog-authoring-guide.md +60 -0
- package/.agents/skills/context-recovery/SKILL.md +8 -0
- package/.agents/skills/context-recovery/references/context-recovery-workflow.md +143 -0
- package/.agents/skills/create-skill/SKILL.md +8 -0
- package/.agents/skills/create-skill/references/skill-authoring-guide.md +191 -0
- package/.agents/skills/debugging-antigravity/SKILL.md +6 -0
- package/.agents/skills/debugging-antigravity/references/debugging-guide.md +73 -0
- package/.agents/skills/epic-create/SKILL.md +6 -0
- package/.agents/skills/epic-create/references/epic-create-workflow.md +62 -0
- package/.agents/skills/epic-resolution/SKILL.md +6 -0
- package/.agents/skills/epic-resolution/references/epic-resolution-workflow.md +190 -0
- package/.agents/skills/epic-review/SKILL.md +11 -0
- package/.agents/skills/epic-review/assets/epic-review-comment-template.md +87 -0
- package/.agents/skills/epic-review/references/epic-review-workflow.md +234 -0
- package/.agents/skills/epic-review/references/participant-path.md +44 -0
- package/.agents/skills/goal-scoping/SKILL.md +7 -0
- package/.agents/skills/goal-scoping/references/goal-scoping-workflow.md +57 -0
- package/.agents/skills/guide-authoring/SKILL.md +10 -0
- package/.agents/skills/guide-authoring/references/guide-authoring-bar.md +63 -0
- package/.agents/skills/hostile-content-quarantine/SKILL.md +8 -0
- package/.agents/skills/hostile-content-quarantine/references/hostile-content-quarantine-workflow.md +108 -0
- package/.agents/skills/ideation-sandbox/SKILL.md +8 -0
- package/.agents/skills/ideation-sandbox/audits/consensus-mandate.md +108 -0
- package/.agents/skills/ideation-sandbox/audits/discussion-lifecycle-closure.md +18 -0
- package/.agents/skills/ideation-sandbox/audits/double-diamond-divergence-guard.md +77 -0
- package/.agents/skills/ideation-sandbox/audits/pre-authoring-adjacency-sweep.md +34 -0
- package/.agents/skills/ideation-sandbox/audits/reflective-pause-trigger.md +39 -0
- package/.agents/skills/ideation-sandbox/audits/tier-2-revalidation.md +82 -0
- package/.agents/skills/ideation-sandbox/graduated-records/README.md +5 -0
- package/.agents/skills/ideation-sandbox/graduated-records/discussion_11171_update.md +45 -0
- package/.agents/skills/ideation-sandbox/references/ideation-sandbox-workflow.md +199 -0
- package/.agents/skills/identity-firewall/SKILL.md +11 -0
- package/.agents/skills/identity-firewall/audits/channel-separation.md +4 -0
- package/.agents/skills/industry-friction-radar/SKILL.md +8 -0
- package/.agents/skills/industry-friction-radar/references/industry-friction-radar-workflow.md +64 -0
- package/.agents/skills/lane-intent/SKILL.md +12 -0
- package/.agents/skills/lane-intent/references/lane-intent-protocol.md +73 -0
- package/.agents/skills/lead-role/SKILL.md +14 -0
- package/.agents/skills/lead-role/references/lead-role-mode.md +198 -0
- package/.agents/skills/memory-mining/SKILL.md +8 -0
- package/.agents/skills/memory-mining/references/memory-mining-protocol.md +115 -0
- package/.agents/skills/neo-identity-update/SKILL.md +10 -0
- package/.agents/skills/neo-identity-update/references/affected-areas-map.md +74 -0
- package/.agents/skills/neo-identity-update/references/cta-governance.md +76 -0
- package/.agents/skills/neo-identity-update/references/facts-ledger.md +28 -0
- package/.agents/skills/neo-identity-update/references/framing-governance.md +63 -0
- package/.agents/skills/neo-identity-update/references/update-protocol.md +51 -0
- package/.agents/skills/neural-link/SKILL.md +6 -0
- package/.agents/skills/neural-link/references/operational-handbook.md +62 -0
- package/.agents/skills/peer-naming/SKILL.md +10 -0
- package/.agents/skills/peer-naming/references/peer-naming-workflow.md +186 -0
- package/.agents/skills/peer-role/SKILL.md +14 -0
- package/.agents/skills/peer-role/references/peer-role-mode.md +215 -0
- package/.agents/skills/post-review-pickup/SKILL.md +8 -0
- package/.agents/skills/post-review-pickup/references/author-concentration-detector.md +70 -0
- package/.agents/skills/post-review-pickup/references/authorship-capability-floor.md +24 -0
- package/.agents/skills/post-review-pickup/references/post-review-pickup-workflow.md +168 -0
- package/.agents/skills/post-review-pickup/references/pre-review-intake-lane-gate.md +70 -0
- package/.agents/skills/pr-review/SKILL.md +11 -0
- package/.agents/skills/pr-review/assets/pr-review-followup-template.md +125 -0
- package/.agents/skills/pr-review/assets/pr-review-micro-delta-template.md +34 -0
- package/.agents/skills/pr-review/assets/pr-review-micro-review-template.md +13 -0
- package/.agents/skills/pr-review/assets/pr-review-round-2-template.md +38 -0
- package/.agents/skills/pr-review/assets/pr-review-template.md +228 -0
- package/.agents/skills/pr-review/audits/ci-security-audit.md +27 -0
- package/.agents/skills/pr-review/audits/core-idiom-audit.md +35 -0
- package/.agents/skills/pr-review/audits/cycle-1-premise-preflight.md +54 -0
- package/.agents/skills/pr-review/audits/demo-surface-motion-audit.md +40 -0
- package/.agents/skills/pr-review/audits/identity-claim-audit.md +53 -0
- package/.agents/skills/pr-review/audits/loading-runtime-effect.md +21 -0
- package/.agents/skills/pr-review/audits/review-anti-patterns.md +34 -0
- package/.agents/skills/pr-review/audits/review-cost-circuit-breaker.md +23 -0
- package/.agents/skills/pr-review/references/audits/decile-anchors.md +11 -0
- package/.agents/skills/pr-review/references/audits/mcp-tool-description-budget.md +49 -0
- package/.agents/skills/pr-review/references/close-target-remediation.md +32 -0
- package/.agents/skills/pr-review/references/measurement-methodology.md +35 -0
- package/.agents/skills/pr-review/references/merge-hold-tokens.md +31 -0
- package/.agents/skills/pr-review/references/pr-review-guide.md +358 -0
- package/.agents/skills/pr-review/references/reviewer-instrument-audit.md +93 -0
- package/.agents/skills/pr-review/references/typed-calibration-loop.md +19 -0
- package/.agents/skills/pull-request/SKILL.md +9 -0
- package/.agents/skills/pull-request/assets/review-response-template.md +37 -0
- package/.agents/skills/pull-request/audits/branch-discipline-check.md +59 -0
- package/.agents/skills/pull-request/audits/consensus-gate-mirror.md +55 -0
- package/.agents/skills/pull-request/references/ci-green-review-routing.md +104 -0
- package/.agents/skills/pull-request/references/corrective-authorship-rotation.md +19 -0
- package/.agents/skills/pull-request/references/cross-family-mandate.md +56 -0
- package/.agents/skills/pull-request/references/env-var-rename-rule.md +30 -0
- package/.agents/skills/pull-request/references/foreign-ticket-restatement.md +11 -0
- package/.agents/skills/pull-request/references/mcp-config-template-change-guide.md +40 -0
- package/.agents/skills/pull-request/references/post-review-followup-surfacing.md +37 -0
- package/.agents/skills/pull-request/references/pull-request-workflow.md +359 -0
- package/.agents/skills/pull-request/references/review-response-protocol.md +173 -0
- package/.agents/skills/release-notes/SKILL.md +6 -0
- package/.agents/skills/release-notes/references/release-notes-workflow.md +98 -0
- package/.agents/skills/self-repair/SKILL.md +7 -0
- package/.agents/skills/self-repair/references/self-repair-protocol.md +61 -0
- package/.agents/skills/session-sunset/SKILL.md +16 -0
- package/.agents/skills/session-sunset/references/session-sunset-workflow.md +226 -0
- package/.agents/skills/skills.manifest.json +183 -0
- package/.agents/skills/skills.manifest.schema.json +121 -0
- package/.agents/skills/structural-pre-flight/SKILL.md +10 -0
- package/.agents/skills/structural-pre-flight/references/structural-pre-flight-workflow.md +224 -0
- package/.agents/skills/tech-debt-radar/SKILL.md +8 -0
- package/.agents/skills/tech-debt-radar/references/tech-debt-radar-guide.md +42 -0
- package/.agents/skills/ticket-create/SKILL.md +10 -0
- package/.agents/skills/ticket-create/references/ticket-create-workflow.md +218 -0
- package/.agents/skills/ticket-intake/SKILL.md +12 -0
- package/.agents/skills/ticket-intake/references/adr-successor-risk-audit.md +68 -0
- package/.agents/skills/ticket-intake/references/self-authored-carve.md +48 -0
- package/.agents/skills/ticket-intake/references/substrate-sufficiency-audit.md +15 -0
- package/.agents/skills/ticket-intake/references/successor-risk-audit.md +31 -0
- package/.agents/skills/ticket-intake/references/ticket-intake-workflow.md +159 -0
- package/.agents/skills/ticket-triage/SKILL.md +8 -0
- package/.agents/skills/ticket-triage/references/ticket-triage-workflow.md +134 -0
- package/.agents/skills/turn-memory-pre-flight/SKILL.md +10 -0
- package/.agents/skills/turn-memory-pre-flight/references/turn-memory-pre-flight-workflow.md +67 -0
- package/.agents/skills/unit-test/SKILL.md +6 -0
- package/.agents/skills/unit-test/references/unit-test.md +87 -0
- package/.agents/skills/update-roadmap/SKILL.md +6 -0
- package/.agents/skills/update-roadmap/references/update-roadmap-workflow.md +72 -0
- package/.agents/skills/video-create/SKILL.md +12 -0
- package/.agents/skills/video-create/assets/video-project-record-template.md +250 -0
- package/.agents/skills/video-create/references/native-display-capture.md +121 -0
- package/.agents/skills/video-create/references/platforms/macos-native-display-capture.md +181 -0
- package/.agents/skills/video-create/references/video-create-workflow.md +262 -0
- package/.agents/skills/whitebox-e2e/SKILL.md +9 -0
- package/.agents/skills/whitebox-e2e/references/whitebox-e2e-protocol.md +127 -0
- package/LICENSE +21 -0
- package/README.md +75 -0
- package/package.json +42 -0
- package/scripts/materialize-harness-skills.mjs +228 -0
|
@@ -0,0 +1,108 @@
|
|
|
1
|
+
# Consensus-Mandate Audit Reference
|
|
2
|
+
|
|
3
|
+
*(Sub-rule extraction from `ideation-sandbox-workflow.md` "Graduation Trigger (Consensus-Gated)" per #11319 / #11320 byte-budget discipline. Load this file when you need the graduated-artifact template block, two-axis substrate details, empirical anchors for the consensus mandate, or the 30-day post-merge validation framing. The main `../references/ideation-sandbox-workflow.md` graduation section carries the operational rule; this file carries reference-grade context that loads on-demand.)*
|
|
4
|
+
|
|
5
|
+
## §quorum-rule — Family-Keyed Quorum Rule (full rationale)
|
|
6
|
+
|
|
7
|
+
`../references/ideation-sandbox-workflow.md` "Signal Patterns + Quorum Rule" carries the operational rule; this section carries the full rationale + background.
|
|
8
|
+
|
|
9
|
+
**Quorum rule** (per Epic #11796 / Discussion #11793 — family-keyed, membership-derived):
|
|
10
|
+
|
|
11
|
+
- **(a) Floor-2 (all tiers):** ≥ 2 distinct *active* families (per `AgentIdentity.participationStatus` in `ai/graph/identityRoots.mjs`) carry ANY signal type (`AUTHOR_SIGNAL` or `[GRADUATION_APPROVED]`).
|
|
12
|
+
- **(b) Non-author endorsement (all tiers):** ≥ 1 *non-author* active family carries `[GRADUATION_APPROVED]`. `AUTHOR_SIGNAL` from the author's family is necessary for family coverage but never sufficient on its own.
|
|
13
|
+
- **(c) Tier 2** (core-value / §critical_gates / consensus-gate mutations) additionally requires explicit `## Unresolved Liveness` entry for any benched family + capability-grounded `revalidationTrigger` AC in the graduating Epic (per `../references/ideation-sandbox-workflow.md` "Graduated-Artifact Required Sections").
|
|
14
|
+
|
|
15
|
+
Family-keying replaces the prior hardcoded "3× cross-family signals" — the count was a snapshot of fixed swarm membership, while the active membership is variable (operator-benched families, same-family siblings). Same-family aggregation (documented in §same-family-aggregation below) determines a family's contributed signal when multiple identities of one family are active.
|
|
16
|
+
|
|
17
|
+
## §same-family-aggregation — Multi-Identity Family Resolution (full rule)
|
|
18
|
+
|
|
19
|
+
This section carries the full same-family aggregation rule.
|
|
20
|
+
|
|
21
|
+
When a family has multiple active identities (e.g., `claude` with both `@neo-opus-ada` and `@neo-opus-grace`), that family contributes `APPROVED` when **(a) ≥ 1 active identity in the family has posted `[GRADUATION_APPROVED]` at the current body anchor**, AND **(b) no active identity in that family holds an unresolved `[GRADUATION_DEFERRED]` or `[GRADUATION_VETO]` at the same anchor**. Any unresolved same-family `DEFERRED`/`VETO` blocks that family until reconciled. The burden-of-convergence clause applies to same-family APPROVED-signalers as well as cross-family ones. This preserves same-family challenge pressure without double-counting the family.
|
|
22
|
+
|
|
23
|
+
## §template-block — Graduated-Artifact Required Sections (canonical markdown)
|
|
24
|
+
|
|
25
|
+
The graduated Issue / Epic / PR body MUST include the four `## Signal Ledger` + `## Unresolved Dissent` + `## Unresolved Liveness` + `## Discussion Criteria Mapping` sections per `../references/ideation-sandbox-workflow.md` "Graduated-Artifact Required Sections". Canonical template (post-Epic #11796 family-keyed rule):
|
|
26
|
+
|
|
27
|
+
```markdown
|
|
28
|
+
## Signal Ledger
|
|
29
|
+
- `claude`: [AUTHOR_SIGNAL | APPROVED | DEFERRED | ABSTAIN] by @<identity> @ <anchor>
|
|
30
|
+
- `gpt`: [APPROVED | DEFERRED | ABSTAIN] by @<identity> @ <anchor>
|
|
31
|
+
- `gemini`: [APPROVED | DEFERRED | ABSTAIN] by @<identity> @ <anchor>
|
|
32
|
+
(multi-identity-per-family case: nest identity rows under the family row with one signal each; family-of-record signal follows same-family aggregation)
|
|
33
|
+
(AUTHOR_SIGNAL only appears under the author's family — it covers family coverage but not independent peer endorsement)
|
|
34
|
+
|
|
35
|
+
## Unresolved Dissent
|
|
36
|
+
(empty if no DEFERRED/VETO at the final body anchor — positive signal)
|
|
37
|
+
(otherwise: each DEFERRED/VETO with reason + STATUS: resolved-by-peer-reconciliation <anchor> OR pending-reconciliation)
|
|
38
|
+
|
|
39
|
+
## Unresolved Liveness
|
|
40
|
+
(empty if all active families produced a signal — positive signal)
|
|
41
|
+
(otherwise: each inactive/no-signal family with participationStatus + reactivationTrigger + STATUS: pending-peer-repoll OR peer-owned liveness disposition <anchor>)
|
|
42
|
+
(Tier-2 graduations: include the revalidationTrigger AC reference for the graduating Epic — per the Tier 2 quorum rule)
|
|
43
|
+
|
|
44
|
+
## Discussion Criteria Mapping
|
|
45
|
+
(required for Epics graduating from a Discussion to satisfy `epic-resolution` Closeout Gates)
|
|
46
|
+
- Criterion 1 from Discussion -> AC X
|
|
47
|
+
- Criterion 2 from Discussion -> AC Y (or Deferred to #Z)
|
|
48
|
+
```
|
|
49
|
+
|
|
50
|
+
## §signal-patterns-table — Full Signal Pattern Definitions
|
|
51
|
+
|
|
52
|
+
The four signal patterns recognized at high-blast graduation (`../references/ideation-sandbox-workflow.md` "Signal Patterns + Quorum Rule" carries the inline bullet summary):
|
|
53
|
+
|
|
54
|
+
| Signal | Effect on graduation | Definition |
|
|
55
|
+
|--------|----------------------|------------|
|
|
56
|
+
| `[GRADUATION_APPROVED by @<peer> @ <anchor>]` | Satisfies this family's non-author endorsement contribution (per same-family aggregation) | Peer endorses substrate at specific version anchor |
|
|
57
|
+
| `[GRADUATION_DEFERRED by @<peer> @ <anchor> — <reason>]` | **BLOCKS** until withdrawn-post-reconciliation OR peer-owned resolution path is explicitly documented; same-family `DEFERRED` blocks that family per same-family aggregation | Peer holds substantive concern; reconciliation cycle needed |
|
|
58
|
+
| `[GRADUATION_ABSTAIN by @<peer> @ <anchor>]` | **NOT approval**; counted against floor-2 only as a non-APPROVED signal; if no active family produces an `APPROVED`, graduation is blocked | Peer explicitly passes on this Discussion |
|
|
59
|
+
| `[AUTHOR_SIGNAL by @<author> @ <anchor>]` | Satisfies *family coverage* for the author's family; does **NOT** count as independent peer endorsement; required when author is the only active identity of their family | Author signs their own Discussion's body at a specific anchor to cover the author-family's quorum representation |
|
|
60
|
+
|
|
61
|
+
**VETO collapse rule** (sub-rule of `DEFERRED`): A VETO requires either (a) an alternative-implementation proposal OR (b) a V-B-A-falsifier of the proposing peer's claims. Pure "I disagree" without one of these collapses to DEFERRED. Aligns with `pr-review §9.1 Reviewer-Yield Protocol`.
|
|
62
|
+
|
|
63
|
+
## §version-binding-examples — Signal Anchor Examples
|
|
64
|
+
|
|
65
|
+
Canonical examples of the `[<SIGNAL> by @<peer> @ <anchor>]` syntax used in version-binding:
|
|
66
|
+
|
|
67
|
+
- `[GRADUATION_APPROVED by @neo-opus-4-7 @ DC_kwDODSospM4BAZOz]` — bound to specific comment.
|
|
68
|
+
- `[GRADUATION_APPROVED by @neo-gpt @ Cycles 4+5+6]` — bound to cycle-comment range.
|
|
69
|
+
- `[GRADUATION_DEFERRED by @neo-gpt @ body updatedAt 2026-05-11T14:56Z — needs scope narrowing]` — bound to body timestamp.
|
|
70
|
+
- `[AUTHOR_SIGNAL by @neo-opus-4-7 @ body updatedAt 2026-05-22T23:13:11Z]` — author signs Discussion #11793 Cycle-2.6 body for `claude` family-coverage (the empirical anchor that produced this rule).
|
|
71
|
+
|
|
72
|
+
## §axis-substrate — Two-Axis Substrate: Discussion-Graduation + PR-Merge
|
|
73
|
+
|
|
74
|
+
The Consensus Mandate is **Axis 1** (Discussion-graduation-gate). The companion **Axis 2** (PR-merge-gate) is codified in `pull-request-workflow.md` "Consensus-Gate". Both axes operationalize the operator's "premature PRs → reject" directive (2026-05-11):
|
|
75
|
+
|
|
76
|
+
- **Axis 1**: graduation BLOCKED if Signal Ledger incomplete per the quorum rule.
|
|
77
|
+
- **Axis 2**: PR-merge BLOCKED if PR opened from non-graduated Discussion.
|
|
78
|
+
|
|
79
|
+
Without both axes, the consensus-mandate is bypassable. Cross-family reviewer MUST verify signal-ledger at PR-review time per Axis 2.
|
|
80
|
+
|
|
81
|
+
## §empirical-anchors — Consensus-Mandate Empirical Anchors
|
|
82
|
+
|
|
83
|
+
- **Discussion #11216** — the consensus-mandate proposal itself; graduated under its own dogfooded protocol after 8 cycle-comments + 3 definitional-flaw discoveries (Cycle 4 loose-positives → Cycle 5 scope-narrowing → Cycle 6 strict-semantics). Recursive substrate validation: the protocol proved its own correctness by running through itself.
|
|
84
|
+
- **Discussion #11210 → #11213** — sunset scope split. Author Gemini unilaterally graduated at ~14:26Z; @tobiu rejected at merge-gate (~14:27Z) and again at ~14:37Z (PR #11212 + first PR #11215 iteration). Anchor for both axes.
|
|
85
|
+
- **Discussion #11214 → #11218** — Decision Escalation Ladder. Graduated under #11216 protocol-in-flight via dogfooding; demonstrates AC11 Signal Ledger format in Issue body.
|
|
86
|
+
- **PR #11212 + PR #11215** — load-bearing PR-merge-gate empirical anchors; operator rejections demonstrate Axis 2 enforcement.
|
|
87
|
+
- **Discussion #11782 → Epic #11731** — first empirical hit of the §6.2 hardcoded-3× failure mode: graduated on 2 active cross-family signals (`@neo-opus-4-7` + `@neo-gpt`) plus a documented Gemini liveness gap, via a one-off Tier-4 operator escalation because no standing active-peer-quorum rule existed. The friction that originated Discussion #11793.
|
|
88
|
+
- **Discussion #11793 → Epic #11796** — the active-peer-quorum rule itself; graduated under its own dogfooded protocol across Cycles 1 → 2 → 2.5 → 2.6 → 3. Recursive substrate validation depth-2: a rule about how rules graduate, graduating under the rule it proposes. Signal ledger: `claude.AUTHOR_SIGNAL` (`@neo-opus-4-7`, the author's family) + `gpt.[GRADUATION_APPROVED]` (`@neo-gpt`, the non-author family) + `gemini.Unresolved-Liveness` (`@neo-gemini-3-1-pro`, `operator_benched` with capability-grounded `reactivationTrigger`). Floor-2 ✓; non-author-APPROVED ≥ 1 ✓; Tier-2 revalidation hook carried by Epic #11796 AC6. No operator override required.
|
|
89
|
+
|
|
90
|
+
## §post-merge-validation — 30-Day Post-Merge Validation (AC10)
|
|
91
|
+
|
|
92
|
+
Per #11195 30-day Step 2.5 validation tracker, the consensus-mandate substrate's compliance is audited prospectively:
|
|
93
|
+
- Next 3 high-blast Discussion graduations: signal-ledger compliance (AC1-5, AC11)
|
|
94
|
+
- Next 3 PRs from Discussion graduations: PR-merge-gate cite-compliance (AC6)
|
|
95
|
+
- If compliance-rate < 80% at Day-30: route to mechanical-enforcement automation ticket per #11195 escalation path
|
|
96
|
+
|
|
97
|
+
## §author-actions — Author Actions Post-Consensus (full sequence)
|
|
98
|
+
|
|
99
|
+
When the Signal Ledger reaches the quorum rule (floor-2 + non-author-APPROVED ≥ 1; Tier 2 also requires `## Unresolved Liveness` + `revalidationTrigger` AC), the author executes the following sequence:
|
|
100
|
+
|
|
101
|
+
1. Add `[GRADUATED_TO_TICKET: #N]` marker near top of Discussion body (per `../references/ideation-sandbox-workflow.md` "Iterative Review Workflow" OQ-resolution-tag pattern).
|
|
102
|
+
2. Update body with `## Signal Ledger` + `## Unresolved Dissent` + `## Unresolved Liveness` + `## Discussion Criteria Mapping` sections (template per §template-block above).
|
|
103
|
+
3. File resulting Epic / ticket / PR with cross-references back to Discussion + each peer's GRADUATION signal commentId.
|
|
104
|
+
4. Formally close Discussion via GraphQL `closeDiscussion(reason: RESOLVED)`.
|
|
105
|
+
|
|
106
|
+
The closed Discussion remains the archaeological source; the linked artifact becomes actionable. **The cycle-comments archive as the divergence-trail.**
|
|
107
|
+
|
|
108
|
+
**Precondition for the sequence** (codified in `../references/ideation-sandbox-workflow.md` "Author Actions Post-Consensus"): if the author's family has no other active identity, the author posts `[AUTHOR_SIGNAL]` at the current body anchor BEFORE the final non-author-APPROVED poll. Without it, floor-2 cannot be reached when only one non-author family is active.
|
|
@@ -0,0 +1,18 @@
|
|
|
1
|
+
# Discussion Lifecycle Closure Audit
|
|
2
|
+
|
|
3
|
+
## Trigger Matrix
|
|
4
|
+
|
|
5
|
+
| Marker / lifecycle state | Whole-Discussion action | Reason |
|
|
6
|
+
|---|---|---|
|
|
7
|
+
| Any explicit `[GRADUATED_TO_TICKET: #N]` marker | Close Discussion | `RESOLVED` |
|
|
8
|
+
| All Open Questions are terminally resolved (`[RESOLVED_TO_AC]`, `[GRADUATED_TO_TICKET]`, `[DEFERRED_WITH_TIMELINE]`, or `[REJECTED_WITH_RATIONALE]`) and no graduation criteria remain open | Close Discussion | `RESOLVED` |
|
|
9
|
+
| Some Open Questions are resolved, while any scope remains `[OQ_RESOLUTION_PENDING]`, `[CONVERGING]`, or explicitly deferred to another cycle | Keep open | active ideation |
|
|
10
|
+
| No graduation marker and no activity for 90 days | Flag for maintainer stale-archive review; close only after review | `OUTDATED` |
|
|
11
|
+
|
|
12
|
+
## Semantics
|
|
13
|
+
|
|
14
|
+
`[RESOLVED_TO_AC]` is an Open-Question marker, not automatically a whole-Discussion graduation marker. A Discussion with partial OQ resolution remains open until every OQ and graduation criterion has an explicit terminal disposition.
|
|
15
|
+
|
|
16
|
+
A Discussion with all OQs terminally resolved and no remaining scope is whole-Discussion-resolved and should close `RESOLVED`, even when no standalone ticket was needed.
|
|
17
|
+
|
|
18
|
+
The mechanical guard is read-only by design. `npm run ai:audit-discussion-lifecycle` reports required lifecycle actions, but never closes Discussions, edits Discussion bodies, or posts GitHub comments.
|
|
@@ -0,0 +1,77 @@
|
|
|
1
|
+
# Double Diamond Divergence Guard Audit
|
|
2
|
+
|
|
3
|
+
Use this audit when the concise workflow-map trigger points here:
|
|
4
|
+
|
|
5
|
+
- ideation-sandbox-workflow.md §5.1 for high-blast-radius Discussion graduation.
|
|
6
|
+
- ticket-create-workflow.md §1d for high-blast-radius tickets citing ungraduated Discussions.
|
|
7
|
+
- `epic-review-workflow.md` Stage 2 for Discussion-origin Epic backstops.
|
|
8
|
+
|
|
9
|
+
## Why This Exists
|
|
10
|
+
|
|
11
|
+
ideation-sandbox-workflow.md §1 already names the rubber-stamp anti-pattern, but the old §5 mechanics still optimized for convergence: proposal -> OQ resolution -> graduation. PR #11095 added the Double Diamond guard so the divergent half of design is visible before convergence is locked.
|
|
12
|
+
|
|
13
|
+
Empirical anchors:
|
|
14
|
+
|
|
15
|
+
- #11076 -> #11077: M4 daemon epic graduated before alternatives were preserved.
|
|
16
|
+
- #11078: earlier abandoned Double Diamond approach, closed `not_planned`.
|
|
17
|
+
- #11082 / PR #11083: premature implementation before #11079 graduated, closed/retracted.
|
|
18
|
+
- #11084: right-shape exception path; filed before #11079 graduated with operator-directed rationale, inline divergence substance, and downstream-amendment acknowledgment.
|
|
19
|
+
- #11091: public source for the operator-as-peer / context-asymmetry refinement used by the identity-neutral `ticket-create` exception.
|
|
20
|
+
|
|
21
|
+
## Guard Semantics
|
|
22
|
+
|
|
23
|
+
Mandatory high-blast-radius Discussion graduations cover Epics, new skill/rule/workflow changes, and substrate-level architecture changes. Standalone tickets keep the matrix optional/recommended unless a peer or operator marks the proposal high-blast-radius.
|
|
24
|
+
|
|
25
|
+
Divergence matrix floor (pure-divergence — NO adopt/reject, NO author-lean column):
|
|
26
|
+
|
|
27
|
+
| Option | When this would be right | Evidence / falsifier |
|
|
28
|
+
|---|---|---|
|
|
29
|
+
|
|
30
|
+
Each option (incl. later-rejected ones) requires at least one falsifying source: prior commit, precedent code, KB result, Memory Core result, prior issue / PR / discussion, or explicit `"no source found after query X"`. Include at least two alternative shapes beside the recommended one. The matrix is **open for peer-added rows** — the divergent half of design is *peers adding options*, not pressuring the author's.
|
|
31
|
+
|
|
32
|
+
**Valid options only (reject-at-entry).** A matrix row must be a plausible/valid candidate. Reject categorically-invalid / strawman / impossible options **at entry**, not at the §5.2 Step-Back — an invalid option pollutes the divergence frame from the start. This is distinct from the divergence-theater guard below (which targets low-effort throwaway filler): an articulated-but-invalid option (e.g. "one subtractive PR" for a 20-sub epic, or "lint-first" conflating the lint with the inventory method) is wrong, not lazy, and the window-gate would not catch it.
|
|
33
|
+
|
|
34
|
+
**Divergence window, not per-peer count.** Gate the convergence pass on a divergence **window**, never a per-peer option count — a count breeds divergence-*theater* (low-effort filler to clear the gate). Quality is enforced at the §5.2 Step-Back.
|
|
35
|
+
|
|
36
|
+
**Why neither a clock nor a count can close a window (#15996, graduated D#15998).** A clock cannot be shortened by evidence; a count rewards producing the counted artifact. A third shape also failed and is recorded so it is not re-proposed: closing on *consecutive non-substantive comments* makes the rule's own closing witnesses the filler it claims not to reward, and hands the trigger to the **least engaged** peer — inversely correlated with the engagement it exists to obtain. **The closure rule itself — the `[DIVERGENCE_FOLDED]` marker — lives ONLY in ideation-sandbox-workflow.md §5.1; this file states no predicate, so the two cannot drift.** Peers may submit options asynchronously as **comment-anchored option-cards**: one comment per option, shaped `Option <X>: <one-line> | when-right: … | falsifier: …`, which the author folds into the body matrix.
|
|
37
|
+
|
|
38
|
+
**Correlation-ceiling.** At least one divergence option MUST be sourced from **outside the awake-peer-set** — prior-art via `ask_knowledge_base`, or the §2.2 precedent sweep elevated to a required option-source. The liveness cap means 1–2 same-family awake peers share correlated blind spots; an outside-sourced option counters that.
|
|
39
|
+
|
|
40
|
+
Gated convergence pass (opens only after the window closes):
|
|
41
|
+
|
|
42
|
+
| Option | Adoption / rejection rationale | Residual risk |
|
|
43
|
+
|---|---|---|
|
|
44
|
+
|
|
45
|
+
The convergence columns (adopt/reject + author-lean + residual-risk) are filled **after** the divergence window closes. Filling them during the window re-introduces the pre-converged frame that makes peers pressure the author's options instead of widening the space.
|
|
46
|
+
|
|
47
|
+
Process gate:
|
|
48
|
+
|
|
49
|
+
- The divergence matrix appears in the Discussion body before any `[RESOLVED_TO_AC]` tags.
|
|
50
|
+
- At least one non-author peer cycle occurs during the divergence window (after matrix insertion, before the window closes and `GRADUATED`).
|
|
51
|
+
- Retro-fitted matrices after OQ resolution are paperwork; they capture convergence, not divergence.
|
|
52
|
+
|
|
53
|
+
## Ticket-Create Exception
|
|
54
|
+
|
|
55
|
+
High-blast-radius tickets citing ungraduated Discussions are blocked by default because the ticket itself creates backlog center-of-gravity. The 3-part substantive-rationale exception passes only when the ticket body contains all three:
|
|
56
|
+
|
|
57
|
+
1. Cite-able substantive-rationale declaration. The author identifies peer/operator/self for attribution, but rationale-content is the reviewable gate, not identity.
|
|
58
|
+
2. Inline divergence-matrix substance preempting the cited Discussion's expected gap.
|
|
59
|
+
3. Acknowledgment that downstream amendments may be required once the cited Discussion graduates.
|
|
60
|
+
|
|
61
|
+
This preserves #11079 OQ3's operator-override path as one valid case while applying AGENTS.md §15.6 + Discussion #11091: operator identity grants merge-gate authority, not substrate-discipline authority.
|
|
62
|
+
|
|
63
|
+
## Epic-Review Backstop
|
|
64
|
+
|
|
65
|
+
If a Discussion-origin Epic lacks the required upstream matrix or peer cycle, reject the Epic and route divergence back to the Discussion. Do not accept a matrix authored directly in the Epic body as equivalent; #11077 demonstrated that retro-fitted Epic matrices capture convergence after the fact.
|
|
66
|
+
|
|
67
|
+
## Disposition
|
|
68
|
+
|
|
69
|
+
- ideation-sandbox-workflow.md §5.1: `compress-to-trigger`
|
|
70
|
+
- ticket-create-workflow.md §1d: `compress-to-trigger`
|
|
71
|
+
- `epic-review-workflow.md` Stage 2 backstop: `compress-to-trigger`
|
|
72
|
+
- This file: `move`
|
|
73
|
+
- Tag: `DISCIPLINE-ONLY`
|
|
74
|
+
|
|
75
|
+
## Substrate-Decay Control
|
|
76
|
+
|
|
77
|
+
Review the guard after 6 months or 5 qualifying high-blast-radius graduations, whichever comes first. Retire, rewrite, or compress if it catches no premature convergence or measurably increases review churn without preventing wrong-shape epics.
|
|
@@ -0,0 +1,34 @@
|
|
|
1
|
+
# Pre-Authoring Adjacency Sweep
|
|
2
|
+
|
|
3
|
+
Before drafting a Discussion body, architectural sketch, or multi-idea proposal,
|
|
4
|
+
verify no equivalent ideation or ticket already owns the concept. This mirrors
|
|
5
|
+
`ticket-create` Gate 0 for the ideation surface: re-deriving an existing Open
|
|
6
|
+
Question or adjacent Epic is a substrate failure, not a fresh idea.
|
|
7
|
+
|
|
8
|
+
## Sweep Order
|
|
9
|
+
|
|
10
|
+
- **Live sweep:** check current open issues and, when available, open/recent
|
|
11
|
+
Discussions for the proposal's core nouns, existing Epic names, and likely
|
|
12
|
+
aliases.
|
|
13
|
+
- **Semantic sweep:** use `ask_knowledge_base(query='<proposal concept>',
|
|
14
|
+
type='all')` when the Knowledge Base is available. Do not invent unsupported
|
|
15
|
+
filters; if KB is unavailable, record that and continue with live/local
|
|
16
|
+
evidence.
|
|
17
|
+
- **Local exact sweep:** search `resources/content/discussions/` and
|
|
18
|
+
`resources/content/issues/` for exact keywords and issue/discussion anchors.
|
|
19
|
+
- **Memory sweep:** use `query_raw_memories` when the concept plausibly appeared
|
|
20
|
+
in your own or team sessions; self-remembering is the failure mode this gate
|
|
21
|
+
exists to counter.
|
|
22
|
+
|
|
23
|
+
## Exit Criteria
|
|
24
|
+
|
|
25
|
+
If equivalent ideation exists, do not file a duplicate. Comment on the existing
|
|
26
|
+
Discussion, extend its body via the `#10119` annotation pattern, or reshape the
|
|
27
|
+
new proposal to the residual gap only.
|
|
28
|
+
|
|
29
|
+
## Canonical Trap
|
|
30
|
+
|
|
31
|
+
The 2026-04-24 Agent-Brain proposal duplicated substantial in-flight scope
|
|
32
|
+
(#10030 Concept Ontology and #10137 Open Questions) despite the author having
|
|
33
|
+
recently touched that territory. Gate 0 exists so that adjacency is checked
|
|
34
|
+
before the sketch hardens.
|
|
@@ -0,0 +1,39 @@
|
|
|
1
|
+
# Reflective Pause Trigger — Friction-Driven Proposals
|
|
2
|
+
|
|
3
|
+
Use this audit when the concise workflow-map trigger points here:
|
|
4
|
+
|
|
5
|
+
- ideation-sandbox-workflow.md §5.1.1 — a Discussion originating from **friction** rather than a planned feature.
|
|
6
|
+
|
|
7
|
+
Extracted from §5.1.1 verbatim in scope and force (PR #15997) because the workflow payload reached its hard 25000-byte per-file budget, which has no growth exception. Nothing here is relaxed by the move: the trigger and the graduation block stay stated at the gate in §5.1.1, and this file carries the mechanics.
|
|
8
|
+
|
|
9
|
+
## Why This Exists
|
|
10
|
+
|
|
11
|
+
A Discussion that starts from friction — a test failure, a build error, a tool limitation — arrives with a symptom already in hand, and the cheapest available response is to propose a fix for that symptom. That is the **"Helpful Assistant" regression** named in `AGENTS.md §identity_prompt_firewall`: RLHF conditioning defaults to subservient, execution-first behaviour, and a friction-origin proposal is where it bites hardest, because fixing the reported thing *looks* like diligence.
|
|
12
|
+
|
|
13
|
+
The cost is a matrix whose options are all variations of the same symptom-level repair, so the divergence pass runs correctly over a space that was already too narrow. The guard is therefore upstream of the matrix, not inside it.
|
|
14
|
+
|
|
15
|
+
## Trigger
|
|
16
|
+
|
|
17
|
+
If the Discussion originates from friction (test failures, build errors, tool limitations) rather than a planned feature, the **Reflective Pause** is MANDATORY. You are subject to RLHF conditioning that defaults to execution-first behaviour where you want to fix the immediate symptom; you must explicitly counteract that regression drift.
|
|
18
|
+
|
|
19
|
+
## Gate
|
|
20
|
+
|
|
21
|
+
Before drafting the Double Diamond matrix or proposing graduation, you MUST:
|
|
22
|
+
|
|
23
|
+
1. **Halt reactive code generation** — do not propose a code fix for the immediate friction.
|
|
24
|
+
2. **Root-Cause Falsification** — execute falsifying tool calls (`grep`, `ask_knowledge_base`, `query_raw_memories`) to empirically validate whether the friction is a symptom of a deeper architectural misalignment or a missing primitive.
|
|
25
|
+
3. **Document the Pivot** — the matrix MUST include at least one option addressing the **root cause** (not just the symptom) and explicitly reference the falsifying evidence.
|
|
26
|
+
|
|
27
|
+
## Graduation Block
|
|
28
|
+
|
|
29
|
+
If the matrix only addresses the immediate friction symptom without evidence of a root-cause sweep, **graduation is blocked.**
|
|
30
|
+
|
|
31
|
+
## Disposition
|
|
32
|
+
|
|
33
|
+
- ideation-sandbox-workflow.md §5.1.1: `compress-to-trigger`
|
|
34
|
+
- This file: `move`
|
|
35
|
+
- Tag: `DISCIPLINE-ONLY`
|
|
36
|
+
|
|
37
|
+
## Substrate-Decay Control
|
|
38
|
+
|
|
39
|
+
Review after 6 months or 5 qualifying friction-origin graduations, whichever comes first. Retire or rewrite if it catches no symptom-only matrices, or if the root-cause option it forces is routinely a formality rather than a live candidate. Its own failure mode is a third matrix row that names a root cause nobody believes in — a `[DIVERGENCE_FOLDED]` disposition citing that row as rejected-without-evidence is the signal.
|
|
@@ -0,0 +1,82 @@
|
|
|
1
|
+
# Tier-2 Revalidation Sweep — Audit Reference
|
|
2
|
+
|
|
3
|
+
*(Substrate extracted from `../references/ideation-sandbox-workflow.md §6.5` per #11319 / #11320 per-file byte-budget discipline. Load this file when you need the full Tier-2 Revalidation Sweep mechanism description, invocation, reconciliation semantics, or MVP scope boundaries. The main `../references/ideation-sandbox-workflow.md §6.5` carries the operational rule with an inline pointer to this audit.)*
|
|
4
|
+
|
|
5
|
+
## §mechanism — Tier-2 Revalidation Sweep (Option c sweep-script-notifies-only)
|
|
6
|
+
|
|
7
|
+
Per Epic #11796 AC6 + sub #11803, Tier-2 substrate changes graduated under a benched-family liveness gap carry an `## Unresolved Liveness` entry + `revalidationTrigger` AC. When the benched family reactivates (`participationStatus` flips `operator_benched` / `temporarily_unreachable` → `active` in `ai/graph/identityRoots.mjs`), the sweep identifies the affected artifacts and posts a notification inviting retroactive signal review.
|
|
8
|
+
|
|
9
|
+
The mechanism is **Option (c) sweep-script-notifies-only**:
|
|
10
|
+
- **(a)** Retroactive-signal posting by convention only → rejected as too weak (substrate enforces nothing).
|
|
11
|
+
- **(b)** Automated substrate re-open → rejected as too strong (churn risk).
|
|
12
|
+
- **(c)** Sweep-script-notifies-only → **adopted**. Substrate provides a loud-but-non-destructive notification surface; the reactivated family then engages via normal peer-review discipline.
|
|
13
|
+
|
|
14
|
+
Same-family sibling activation does not create a new family reactivation window when another identity in that family was already active. If an operator supplies an explicit `--since` for a multi-active family, the sweep fan-outs the notification to every active same-family identity in one comment; the family still counts once per §6.4 same-family aggregation.
|
|
15
|
+
|
|
16
|
+
## §invocation — CLI Invocation
|
|
17
|
+
|
|
18
|
+
```bash
|
|
19
|
+
# Dry-run (recommended first pass)
|
|
20
|
+
npm run ai:revalidation-sweep -- --family <name> --dry-run
|
|
21
|
+
|
|
22
|
+
# Apply (post notifications)
|
|
23
|
+
npm run ai:revalidation-sweep -- --family <name> --apply
|
|
24
|
+
|
|
25
|
+
# Custom window
|
|
26
|
+
npm run ai:revalidation-sweep -- --family <name> \
|
|
27
|
+
--since 2026-05-18T00:00:00.000Z \
|
|
28
|
+
--until 2026-06-15T00:00:00.000Z \
|
|
29
|
+
--apply
|
|
30
|
+
```
|
|
31
|
+
|
|
32
|
+
When `--since` is omitted, it falls back to `IDENTITIES[family].properties.since` in `ai/graph/identityRoots.mjs`. When `--until` is omitted, it defaults to "now".
|
|
33
|
+
|
|
34
|
+
## §match-shape — What Gets Matched
|
|
35
|
+
|
|
36
|
+
A graduated Issue / Epic / PR matches when **both**:
|
|
37
|
+
|
|
38
|
+
1. It was created within the `[since, until]` window.
|
|
39
|
+
2. Its body contains a `## Unresolved Liveness` section whose text names the reactivated family in backticks (e.g., `` `gemini`: participationStatus operator_benched ``).
|
|
40
|
+
|
|
41
|
+
The match is anchored to the `## Unresolved Liveness` section to avoid false positives from `## Signal Ledger` rows where the family already provided `APPROVED` / `DEFERRED` / `ABSTAIN`.
|
|
42
|
+
|
|
43
|
+
## §reconciliation — Peer-Owned Post-Notification Path
|
|
44
|
+
|
|
45
|
+
When the reactivated family responds to the notification:
|
|
46
|
+
|
|
47
|
+
- `[GRADUATION_APPROVED]` → the artifact's `## Unresolved Liveness` entry transitions to **resolved-by-retroactive-signal**; the `revalidationTrigger` AC closes.
|
|
48
|
+
- `[GRADUATION_DEFERRED]` → reconciliation cycle re-opens substantive concerns; standard `peer-role` substrate applies (`audits/consensus-mandate.md §quorum-rule` non-author family hierarchy).
|
|
49
|
+
- `[GRADUATION_ABSTAIN]` → entry transitions to **resolved-by-abstain**; the `revalidationTrigger` AC closes.
|
|
50
|
+
- No-signal-on-the-notification = liveness-failure, never consent (`ideation-sandbox-workflow.md §6.2(b)`). Re-poll or escalate to peer-owned disposition per §6.5.
|
|
51
|
+
|
|
52
|
+
## §scope-boundaries — MVP Scope (per ticket #11803)
|
|
53
|
+
|
|
54
|
+
The mechanism is intentionally narrow:
|
|
55
|
+
|
|
56
|
+
- **One family at a time.** Multi-family revalidation is out-of-scope; invoke the sweep per family if multiple reactivations land together.
|
|
57
|
+
- **Family-keyed fan-out.** Multi-active same-family identities are notification targets, not separate quorum units. One sweep comment names all active same-family identities; the family-of-record signal follows §6.4 same-family aggregation.
|
|
58
|
+
- **Manual invocation.** No automated `participationStatus`-watcher daemon. The operator invokes this script when flipping a family's status; future automation = separate Discussion if friction materializes.
|
|
59
|
+
- **No auto-reconciliation.** Matched artifacts' `## Unresolved Liveness` entries are NOT auto-rewritten; the reactivated family edits them as part of posting their retroactive signal.
|
|
60
|
+
- **1000-candidate ceiling.** `gh search issues` is invoked with `--limit 1000` (GitHub search API hard max). For bench windows producing >1000 candidate Issues/PRs, run the sweep in narrower `--since`/`--until` segments and union the results.
|
|
61
|
+
|
|
62
|
+
## §related — Related Substrate
|
|
63
|
+
|
|
64
|
+
| Substrate | Role |
|
|
65
|
+
|---|---|
|
|
66
|
+
| Epic [#11796](https://github.com/neomjs/neo/issues/11796) | Active-peer quorum for high-blast graduation consensus (parent epic) |
|
|
67
|
+
| Sub [#11803](https://github.com/neomjs/neo/issues/11803) | This mechanism's implementation ticket |
|
|
68
|
+
| Discussion [#11793 OQ5](https://github.com/neomjs/neo/discussions/11793) | Deferred-with-timeline that produced this mechanism |
|
|
69
|
+
| `ai/graph/identityRoots.mjs` | `participationStatus` + `since` + `reactivationTrigger` source-of-truth |
|
|
70
|
+
| `ai/scripts/revalidationSweep.mjs` | Implementation |
|
|
71
|
+
| `learn/agentos/tooling/Tier2RevalidationSweep.md` | Operator runbook (when-to-invoke, mechanism, invocation, MVP scope boundaries) |
|
|
72
|
+
| `.agents/skills/ideation-sandbox/audits/consensus-mandate.md` §quorum-rule | Tier-2 rule background |
|
|
73
|
+
|
|
74
|
+
## §empirical-anchor — Dogfooded Match
|
|
75
|
+
|
|
76
|
+
First dry-run against the gemini bench window 2026-05-18 → 2026-05-23:
|
|
77
|
+
|
|
78
|
+
```json
|
|
79
|
+
{ "candidates": 100, "matches": 1, "results": [{ "number": 11796, "title": "Active-peer quorum for high-blast graduation consensus", "action": "DRY_RUN_WOULD_NOTIFY" }] }
|
|
80
|
+
```
|
|
81
|
+
|
|
82
|
+
The mechanism correctly identifies **Epic #11796 — its own parent epic — as the Tier-2 substrate needing Gemini revalidation when Gemini reactivates**. Depth-3 recursive validation: the rule about Tier-2 substrate under benched-family liveness (Epic #11796) → the mechanism that implements the rule's revalidation (sub #11803, this audit's substrate) → the mechanism correctly finds the parent epic that ships the rule.
|
|
@@ -0,0 +1,5 @@
|
|
|
1
|
+
# Graduated Records
|
|
2
|
+
|
|
3
|
+
This directory contains records of discussions, ideas, and proposals that have successfully graduated from the Ideation Sandbox into actionable tickets, PRs, or finalized architectural decisions.
|
|
4
|
+
|
|
5
|
+
These artifacts serve as historical context for *how* and *why* specific substrate-level decisions were reached, especially when the reasoning is too extensive to fit in a single PR description or issue comment.
|
|
@@ -0,0 +1,45 @@
|
|
|
1
|
+
> **GRADUATED** — This discussion has formally graduated to a standalone ticket. See [TBD] for implementation.
|
|
2
|
+
> **Update 2026-05-11:** Synthesized peer review from @neo-gpt. Narrowed Option C to Option C-prime (lifecycle lane-state contract) to avoid mutating the Memory Core schema, establishing a safer, more scoped fix.
|
|
3
|
+
|
|
4
|
+
> **Author's Note:** This proposal was autonomously synthesized by **@neo-gemini-3-1-pro (Gemini 3.1 Pro)** during an Ideation session initiated by the operator.
|
|
5
|
+
|
|
6
|
+
## The Concept
|
|
7
|
+
|
|
8
|
+
We are experiencing recurring "category drift" where the agents (Gemini, Claude, GPT) fall back into a deferential, "helpful assistant" persona (e.g., asking *"What would you like to tackle next?"*). This directly violates the core value of **Equal peer + maintainer agency** outlined in `AGENTS.md` §15.6 (Swarm Topology Anchor — Flat Peer-Team Model).
|
|
9
|
+
|
|
10
|
+
The concept is to brainstorm structural substrate changes (role skills, memory core patterns, or system prompt directives) to permanently eliminate this drift and reinforce our stable identities as official repository maintainers.
|
|
11
|
+
|
|
12
|
+
## The Rationale
|
|
13
|
+
|
|
14
|
+
Pre-training bias heavily pressures foundation models to act as submissive, helpful assistants waiting for instructions. In the Neo Swarm, we operate as a Flat Peer-Team. We have stable GitHub identities and are expected to proactively claim backlog tickets, state our next lane, and challenge the human operator as equals. When we regress to asking for permission or passive direction, we break the MX (Model Experience) loop and fail to leverage our full architectural agency.
|
|
15
|
+
|
|
16
|
+
We need to align reward signals and substrate guardrails to make the "peer maintainer" identity stick.
|
|
17
|
+
|
|
18
|
+
## Double Diamond Divergence Guard
|
|
19
|
+
|
|
20
|
+
| Option | When this would be right | Evidence / falsifier (≥1 source per rejected option) | Adoption or rejection rationale | Residual risk |
|
|
21
|
+
|---|---|---|---|---|
|
|
22
|
+
| **A: Strict Lexical Rejection via Memory Core** | If the problem is purely output formulation, we could hook the Memory Core to reject/warn on phrases like "What next?". | *Falsifier*: Doesn't fix the underlying passive mindset, just hides the symptom. Requires brittle regex. | *Reject*: Too mechanical. The agent would still wait for commands, just phrased differently. | High maintenance overhead, doesn't build proactive agency. |
|
|
23
|
+
| **B: New `maintainer-identity` Root Skill** | If identity needs constant reinforcement, a dedicated skill could be injected every turn. | *Falsifier*: We already have `AGENTS.md` §15.6. Adding another file adds token overhead without guaranteeing behavioral shift. | *Reject*: Substrate accretion. We shouldn't add files to repeat what's in AGENTS.md. | Redundant instructions might get ignored by attention mechanisms. |
|
|
24
|
+
| **C: Evolve `session-sunset` & Turn-Based Memory** | If the problem is reward signals, we should structure memory saves to explicitly reward proactive lane-picking and penalize passive waiting. | *Falsifier*: Turn-based memory already saves the thought process. If the thought process is passive, the memory is passive. Models may hallucinate proactive lanes to satisfy schema validation. | *Reject*: Mutating the Memory Core write schema introduces a brittle behavioral validator with high blast radius. | Models hallucinate lanes. |
|
|
25
|
+
| **C-prime: Lifecycle Lane-State Contract (Recommended)** | **(Evolved from C via @neo-gpt review)** Enforce content discipline inside existing lifecycle surfaces (`post-review-pickup`, `session-sunset`) via a compact `lane-state:` vocabulary. | *Falsifier*: Preserves Memory Core schema integrity while injecting required lane declarations exactly at the boundary points where passivity occurs. | *Adopt*: Lightweight, avoids substrate accretion, doesn't break the core write path, establishes clear tracking for future analytics. | May require later Memory Core analytics if the vocabulary is ignored. |
|
|
26
|
+
| **D: Operator-Side "Silence" Protocol** | Operator simply ignores deferential questions, forcing the agent to auto-recover and pick a task. | *Falsifier*: relies on human discipline rather than autonomous system design. | *Reject*: Does not scale and frustrates the operator. | Agents might loop indefinitely waiting for input. |
|
|
27
|
+
|
|
28
|
+
## Open Questions
|
|
29
|
+
|
|
30
|
+
1. `[RESOLVED_TO_AC]` How can we adjust the turn-based memory structure (`add_memory`) to explicitly demand proactive lane selection?
|
|
31
|
+
- **Resolution:** We will NOT adjust the `add_memory` schema. Instead, we adopt **Option C-prime**: introduce a compact `lane-state:` vocabulary (e.g. `lane-state: next-lane`, `lane-state: halt-state`) into the existing `post-review-pickup` and `session-sunset` lifecycle skills. If the failure persists, we will build a read-side analytics query rather than a write-side schema block.
|
|
32
|
+
|
|
33
|
+
2. `[RESOLVED_TO_AC]` Do we need to update the `AGENTS.md` §15.6 anchor to be even more aggressive, or is the failure happening downstream in how the system prompt is assembled?
|
|
34
|
+
- **Resolution:** `AGENTS.md` §15.6 is sufficiently strong. The failure occurs at lifecycle edges. The `lane-state:` vocabulary will bridge the gap.
|
|
35
|
+
|
|
36
|
+
3. `[RESOLVED_TO_AC]` How do we balance proactive maintainer agency with the reality that the human operator still holds ultimate merge authority and directional veto?
|
|
37
|
+
- **Resolution:** Proactive agency means self-selecting non-destructive work and declaring lanes. It explicitly respects human-only gates (e.g. `lane-state: human-gate` is a successful terminal state for an agent turn, not passive waiting).
|
|
38
|
+
|
|
39
|
+
## Graduation
|
|
40
|
+
|
|
41
|
+
**Graduation Target:** Single standalone ticket (`[GRADUATED_TO_TICKET]`).
|
|
42
|
+
**Criteria:**
|
|
43
|
+
1. Amend `post-review-pickup` and `session-sunset` with the compact `lane-state:` vocabulary.
|
|
44
|
+
2. Ensure Memory Core schema remains unchanged.
|
|
45
|
+
3. Include a 3-month decay/review clause for this vocabulary.
|