[ruvnet]

Closes #203.

Hypothesis (frozen before implementation)

Given a run of consecutive non-self-review nights that all resolve to CONSOLIDATE because the review backlog stays over the limit with an eligible draft always available, when a CONSOLIDATE_STREAK cap (7 nights) is added to CONSOLIDATE's entry condition, forcing RESEARCH-ONLY once the cap is reached even with an eligible draft still waiting, then new research resumes within a bounded number of nights regardless of how long the human review backlog takes to clear, with zero regression in the test suite or any compiled target's golden snapshot.

Candidate

packages/compile/src/v2.ts:

• Added export const CONSOLIDATE_STREAK_CAP = 7.
• v2ModeSelection(): added a CONSOLIDATE_STREAK measurement instruction (consecutive non-routine-self-review nights with neither a new <branchPrefix><DATE>-<surface> branch nor a new [Dream Cycle <DATE>] issue — i.e. nights whose only visible action was CONSOLIDATE's refresh-and-comment); added AND CONSOLIDATE_STREAK < 7 to CONSOLIDATE's entry condition; added OR CONSOLIDATE_STREAK >= 7 to RESEARCH-ONLY's entry condition (so a backlog a human hasn't cleared no longer withholds research indefinitely); added CONSOLIDATE_STREAK to the printed MODE line.

packages/compile/src/index.ts: re-exported CONSOLIDATE_STREAK_CAP.

Does not touch implementationTrack.reviewBacklogLimit or selfReview.everyDays (both explicitly off-limits to self-review, per ROUTINE_CONSTITUTION), CONSOLIDATE's never-merge/never-close/never-force-push limits, the independent critic/reward-hack checks, any evaluator entrypoint, witness computation, or ledger schema.

Evaluation Receipt

Real evaluator: npm test (vitest run && node --test scripts/*.test.mjs).

	Baseline (3fc3222d, parent, post-unshallow)	Candidate
vitest	886	887 (+1 new, 0 regressions)
governance	195	195 (unchanged)
lint / typecheck	clean	clean
mirror check (compile-targets.mjs --check)	pass	pass (all 5 targets regenerated + reviewed)

Self-hosted dream.config.json compile output confirmed byte-identical to its regenerated mirror. All 5 targets sharing this compile engine (self-hosted + ruflo/metaharness/ruvector/ruclip) had their mirrors regenerated — same clause, no per-target parameter.

Baseline

Parent commit 3fc3222d2e5566a203331d3686c4d2eaf06713a8 (main), this session's own checkout (unshallowed to resolve the same already-tracked environmental governance-test failure as PR #202/#150).

Darwin Lineage

DARWIN=not-applicable — a MODE-selection threshold/prose fix has no evolvable population or fitness function.

Evidence

OBSERVATION: main unchanged at 3fc3222d since PR #196 (2026-10-19). OBSERVATION: 21 open dream/* PRs, 0 new ones since PR #194 (2026-10-18) — 16 non-self-review nights, all CONSOLIDATE. OBSERVATION: 6 of those nights independently confirmed via PR comments on #161/#180/#182/#184/#189/#192 (2026-10-30→11-04), each a real merge-main-in + fresh critic + marker. INFERENCE: this pattern can recur indefinitely because CONSOLIDATE's condition depends only on backlog size and eligibility, both chronically satisfied. DECISION: cap the streak rather than the limit itself. Full report with the scored prior forecast: docs/dream-cycle/2026-11-05-routine-self-review-report.md.

Reward-Hack Check

Independent critic dispatched (separate-context subagent, no authorship stake, given only the diff, ROUTINE_CONSTITUTION, and told to attack it against the off-limits list, test vacuity, and specification ambiguity): CLEAR. It independently re-ran npm run build -w packages/compile, vitest run on the three affected test files (181 passed), lint, typecheck, and compile-targets.mjs --check (all green); confirmed the CONSOLIDATE_STREAK "new branch OR new issue" discriminator is exhaustive against the real STEP numbering (RESEARCH-ONLY always files an issue per STEPS 15–18; IMPLEMENT/RESEARCH file both; only CONSOLIDATE does neither); confirmed the new test is not vacuous. Two disclosed non-blocking nitpicks: the given live-check command is open-PR-only (an under-count biases toward more research, not toward dodging the cap — not mission-defeating); and the cap value (7) is defensible but a round choice. No test weakened, no benchmark/threshold/gold data touched; one test added, none removed.

Security Review

Compiled-instructions-only change; no new tool, permission, credential, or network call. Reduces a resource/availability risk (new research silently withheld indefinitely while a human-only-clearable backlog persists) rather than introducing one.

Regression Analysis

11 files changed: 2 logic files (v2.ts new constant + template text, index.ts re-export), 1 test file (+1 test block), 2 generated snapshot files, 5 mechanically-regenerated compiled mirrors, 1 new report file. 887 vitest + 195 governance pass; lint/typecheck clean; mirror check clean. No GLOBAL INVARIANT, verdict rule, promotion gate, merge policy, CONSOLIDATE limit, critic/reward-hack check, evaluator entrypoint, witness computation, ledger schema, reviewBacklogLimit/everyDays, research floor, or config-validation/golden-snapshot test touched in substance — confirmed by diff review and the independent critic, as required by the self-review constitution (ADR-0011).

ADR

None — a MODE-selection threshold/prose fix to the routine's own compiled template, not an architectural decision.

Gist

GIST=LOCAL — no gh CLI (GH_TOKEN invalid this session), no gist MCP tool. Full report committed in this PR.

Issue

#203

Witness

Self-verified via dream-machine witness verify-report (canonical-exclusion scheme) — VALID, bound to session commit 3fc3222d2e5566a203331d3686c4d2eaf06713a8. Triple is in the report's own ## Witness section.

Merge Policy

This PR stays draft, regardless of verdict. Per STEP 1.6/ADR-0011's routine constitution: "A self-review PR always stays human-review-only, whatever prMode says." This session never merges and never autonomously promotes candidate state.

---

Generated by Claude Code

[ruvnet]

CI red, not this PR's: dependency-audit (full dev-toolchain npm audit --audit-level=high) fails on the current head (6568339) with 15 findings (14 moderate, 1 high — vite@8.3.0-8.3.2→launch-editor advisory family, plus an eslint/@typescript-eslint toolchain cluster).

This PR's diff never touches package.json/package-lock.json — git diff origin/main -- package.json package-lock.json is empty, so the lockfile is byte-identical to main's. Reproduced locally: npm ci && npm audit --audit-level=high fails identically. Same pre-existing, already-tracked dev-toolchain advisory family as issue #37 (open since 2026-08-27) and issue #197 (opened 2026-10-21), hit by every recent PR sharing this lockfile (#189, #192, among others) — no fix PR exists yet to port (the real fix is a Vitest/Vite/eslint major-version bump, out of scope for this MODE-selection-template fix).

Production-reachable scope (dependency-security, npm audit --omit=dev) is unaffected by this — these are dev-only findings. All other jobs on this head are green or still in progress (test (22), no-optional-deps, CodeQL, Human merge policy guard all passed). Not re-running the job — this is a deterministic advisory-database finding against an unchanged lockfile, not a flake.

---

Generated by Claude Code