feat(orb): salvageability axis — hold fixable, well-authored PRs instead of one-shot-closing

[JSONbored]

Closes #8962. Advances #8828.

The decision audit's deepest finding: defect-confidence measures whether the defect is REAL, but close-correctness depends on whether closing is the right ACTION — 28 audited holds carried real findings at confidence 0.55–0.97 and still merged (17 as-is, 11 after the author fixed the flagged defect), while acted closes at the same confidences ran 96.0% correct. Same confidence, opposite close-correctness, split entirely by salvageability. That class is the road from ~96% to the 99.5% target.

The score (src/review/salvageability.ts, pure): deterministic and explainable — no AI call, every point named in factors:

• defect-class fixability from the finding text (mechanical +45 / unknown +15 / structural +0 — and structural WINS when both match: no reputation makes a fabricated test or scope violation salvageable-in-place);
• the author's realized MERGED history in THIS repo (+40 at ≥3, +25 at ≥1 — outcomes, not open PRs: the feat(orb): weekly stratified decision-audit sampling with a frozen adjudication rubric #8840 lesson);
• in-flight iteration (+15 at ≥2 review cycles).

The wire (salvageability-wire.ts): two cheap reads (merged count via review_audit pr_outcome, cycles via decision_records), fail-OPEN — a DB blip degrades to today's behavior, never blocks or unblocks.

The gate seam (resolveAiReviewSalvageableHold): resolveAiReviewLowConfidenceHold's sibling for the OTHER side of the floor — fires only when the knob is set, the failure is AI-judgment-only, every blocker sits at/above the close-confidence floor, and the score clears gate.aiReview.salvageabilityMinScore. Routes into the existing held-for-review path with fix-it guidance; the low-confidence hold keeps precedence below the floor. Knob absent (default) short-circuits before any IO.

Auditability: decision records bump to schema v3 carrying salvageability: {score, factors} whenever an AI judgment shaped the decision (recomputed at finalize even when the knob is off, so the boundary accrues evidence BEFORE anyone enables it); the adjudication rubric gains the additive salvageable_close reason category (correct/incorrect semantics unchanged — version 1 labels stay comparable).

Config-as-code parity sweep: engine manifest type/parse/serializer/round-trip + src overlay + policy passthrough + OpenAPI schema (regenerated) + both example-YAML twins (the #8846 lesson).

Tests: 12 pure/resolver/wire tests (every factor tier, structural-wins, all stand-down arms, fail-open); manifest parse/round-trip/overlay; policy passthrough both arms; and an end-to-end queue test — an ABOVE-floor consensus defect on a salvageable PR (manifest knob 60, mechanical finding, 2 merged priors) is HELD not closed, with the v3 record carrying score 70 and its factors. Changed-line coverage: all statements and branches covered.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	loopover-ui	bc9e81a	Commit Preview URL Branch Preview URL	Jul 26 2026, 03:39 PM

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[github-actions]

Logic backtest

Replayed 0 historical case(s) for linked_issue_scope_mismatch through the base (904aeb9) and head (bc9e81a) versions of its detection logic (corpus checksum 4f53cda18c2b).

Backtest comparison: linked_issue_scope_mismatch

Verdict: unchanged — no comparable axis moved.

Advisory only — this check never blocks merge (#8105).

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
21993	1	21992	21

View the top 1 failed test(s) by shortest run time

test/unit/miner-worktree-allocator-collisions.test.ts > loopover-miner worktree allocator collisions (#4298) > returns distinct worktree paths when multiple processes acquire simultaneously

Stack Traces | 0.731s run time

AssertionError: expected 1 to be 5 // Object.is equality

- Expected
+ Received

- 5
+ 1

 ❯ test/unit/miner-worktree-allocator-collisions.test.ts:148:41

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

[loopover-orb]

Caution

🛑 LoopOver review result - fixes required

_{Review updated: 2026-07-26 15:34:45 UTC}

19 files · 1 AI reviewer · no blockers · CI failing · unstable

🛑 Suggested Action - Manual Review

Review summary
The AI review returned non-blocking notes for this change but did not include a separate narrative summary. Review the nits below before deciding this PR.

Nits — 5 non-blocking

• src/queue/processors.ts: computeSalvageabilityForTarget is invoked twice per PR when the knob is on and an AI-judgment blocker exists (once for the plan-site hold, once again unconditionally at decision-record finalize) — two extra DB reads that could be avoided by threading the first result through instead of recomputing.
• src/review/salvageability.ts: the MECHANICAL_PATTERN/STRUCTURAL_PATTERN regexes are long and any future addition to either taxonomy requires touching a dense inline pattern — consider extracting the phrase lists into named arrays for easier maintenance.
• src/review/salvageability.ts: the point weights (45/40/25/15/3/2) are explained in prose comments but not as named constants, making the audit-relevant numbers harder to grep for from callers.
• The two `validate*` CI checks failed with no detail provided, and this branch is 6 commits behind the default branch — that's a plausible cause worth ruling out via rebase before assuming a content defect.
• PR author also opened the linked issue — Link an issue that was opened by a different contributor, or provide a rationale for why this self-authored issue represents genuine discovery work.

CI checks failing

• validate
• validate-tests
• validate-code

Decision drivers

• ✅ Code review — No blockers (1 reviewer)
• ✅ Gate result — Passing (No configured blocker found.)

Context & advisory signals — never blocks the verdict

Signal	Result	Evidence
Linked issue	✅ Linked	#8962
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	✅ 20/20	Low review scope from cached public metadata (1 linked issue).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 13 registered-repo PR(s), 13 merged, 293 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 13 PR(s), 293 issue(s).
Improvement	✅ Minor	risk: clean · value: minor

Review context

• Author: JSONbored
• Role context: owner (maintainer lane)
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: Python, TypeScript, Ruby, Go, MDX, Shell, Solidity, JavaScript
• Official Gittensor activity: 13 PR(s), 293 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Start here: Treat this as maintainer-lane context rather than normal contributor-lane activity.
• Then work through the remaining 1 step in the Signals table above.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Change scope = cached public metadata such as size labels, draft state, and review-burden hints.
• Validation posture = whether the PR provides enough public validation/test evidence for maintainer review.
• Contributor workload = public contributor activity and cleanup pressure, not a repo-wide quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

🧪 Chat with LoopOver

Ask LoopOver a question about this PR directly in a comment — grounded only in the same cached, public-safe facts shown above, never a new claim.

• @loopover ask <question> answers contribution-quality Q&A with source citations and freshness.
• @loopover chat <question> answers in natural prose from cached decision-pack facts via local inference (maintainer/collaborator; read-only).
• A plain-language @loopover mention with a real question is routed to the closest matching read-only command automatically — no exact syntax required.

Full command reference: https://loopover.ai/docs/loopover-commands

🧪 Experimental — new and may change.

Decision record

• action: hold · clause: success
• config: 03a7f8b529a9 · pack: oss-anti-slop
• record: e6cf91f658a0 (schema v2, head 2cb4e67)

Visual preview

Route	Viewport	Before (production)	After (this PR's preview)	Diff
/	desktop	before /	after /	—
/	mobile	before / (mobile)	after / (mobile)	—

Click any thumbnail to open the full-size screenshot. Before = production · After = this PR's preview deploy.

Scroll preview

Route	Before (production)	After (this PR's preview)
/	before / (scroll)	after / (scroll)

A short scroll-through clip (desktop) — click either thumbnail to open the full animation. Evidence for scroll-linked behavior a single screenshot can't show.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed}

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by LoopOver, a quiet PR intelligence layer for OSS maintainers.

• Re-run LoopOver review