fix(scoring): guard credibility against non-finite stale and unlinked counts

[dripsmvcp]

Summary

inferCredibility (in src/scoring/preview.ts) derives a contributor credibility multiplier from cached evidence counts. It coerces mergedPullRequests, stalePullRequests, and unlinkedPullRequests with Number(...), but only finiteness-guards merged:

if (!Number.isFinite(merged)) return 0.8;            // guards merged ONLY
return clamp(0.75 + merged * 0.04 - stale * 0.03 - unlinked * 0.02, 0.25, 1);

The evidence payload is a loosely-typed cache (Record<string, JsonValue>), so any of these fields can arrive as a non-numeric cached value (string, object, …). A non-finite stale or unlinked propagates NaN into the arithmetic, and clamp cannot rescue it (Math.min(1, Math.max(0.25, NaN)) is NaN) — so the whole credibility multiplier, and through buildScorePreview the gates.credibilityObserved and scoreEstimate.estimatedMergedScore, become NaN. The identical garbage in mergedPullRequests is already tolerated (it returns the neutral 0.8); only stale/unlinked were left unguarded.

Fix: extend the existing guard to all three counts, so a malformed stale/unlinked degrades to the same neutral 0.8 fallback that a malformed merged already produces, keeping the score finite. Pure helper; no signature or public-surface change.

No linked issue: this is a small, self-evident defensive-correctness fix to one pure scoring helper, fully described above; per CONTRIBUTING.md / .gittensory.yml (linkedIssuePolicy: preferred) a focused fix like this does not require a linked issue.

Scope

• The PR title follows type(scope): short summary Conventional Commit format.
• This PR is focused and does not mix unrelated backend, UI, MCP, docs, dependency, and deploy changes.
• This follows CONTRIBUTING.md and does not reintroduce GitHub Pages, VitePress, site/, or CNAME.
• I linked an issue, or this is small enough that the summary explains why an issue is not needed (no-issue rationale above).

Validation

• git diff --check
• npm run actionlint
• npm run typecheck
• npm run test:coverage locally — both newly guarded branches (stale and unlinked non-finite) and the well-formed arithmetic path are covered; the changed lines are at 100% line and branch coverage and global coverage stays at/above the 97% line/branch threshold.
• npm run test:workers
• npm run build:mcp
• npm run test:mcp-pack
• npm run ui:openapi:check
• npm run ui:lint
• npm run ui:typecheck
• npm run ui:build
• npm audit --audit-level=moderate
• New behavior has unit tests for the new branches and the fallback path (test/unit/scoring.test.ts).

If any required check was skipped, explain why:

• npm run test:ci (which wraps git diff --check, actionlint, db:migrations:check, typecheck, test:coverage, test:workers, build:mcp, test:mcp-pack, ui:openapi:check, ui:version-audit, ui:lint, ui:typecheck, ui:test, ui:build) ran fully green locally.
• npm audit --audit-level=moderate surfaces only one pre-existing transitive advisory (undici) that this PR does not introduce — the diff changes no dependency manifest or lockfile, so the dependency-review job is unaffected (that ambient advisory is tracked by the scheduled audit.yml workflow + Renovate, not by PR CI).

Safety

• No secrets, wallet details, hotkeys, coldkeys, user PATs, private keys, raw trust scores, private rankings, or private maintainer evidence are exposed (numeric-only pure-function change).
• Public GitHub text stays sanitized, low-noise, and does not imply compensation guarantees or optimization tactics (no public-comment surface touched).
• Auth, cookie, CORS, GitHub App, Cloudflare, or session changes include negative-path tests — N/A: no auth/cookie/CORS/GitHub App/Cloudflare/session changes.
• API/OpenAPI/MCP behavior is updated and tested where needed — N/A: no API/OpenAPI/MCP surface change (internal helper only; ui:openapi:check passes unchanged).
• UI changes use live API data or real empty/error/loading states — N/A: no UI changes.
• Public docs/changelogs are updated where needed — N/A: no docs/changelog changes.

Notes

• The added regression test asserts each malformed-count case yields credibilityObserved === 0.8 and a finite estimatedMergedScore, and that well-formed counts still flow through the arithmetic (so the guard does not over-fire).

[superagent-security]

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.00%. Comparing base (7b50835) to head (e2e880e).
⚠️ Report is 2 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1155   +/-   ##
=======================================
  Coverage   95.00%   95.00%           
=======================================
  Files         177      177           
  Lines       19827    19827           
  Branches     7133     7133           
=======================================
  Hits        18837    18837           
  Misses        396      396           
  Partials      594      594           

Files with missing lines	Coverage Δ
src/scoring/preview.ts	98.82% <100.00%> (ø)

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ghost]

Tip

🟩🟩🟩🟩🟩🟩🟩🟩🟩🟩🟩🟩

✅ Gittensory review — safe to merge

2 files · 2 AI reviewers · no blockers · readiness 100/100 · CI green

✅ Approved — safe to merge

Review summary
The change expands the finiteness guard in `inferCredibility` to include stale and unlinked pull request counts, preventing NaN propagation and ensuring a fallback credibility of 0.8. A new test validates this behavior for malformed stale and unlinked values. All existing tests pass, confirming functional correctness.

Nits (5)

• The added comment in `preview.ts` is quite long and could be split for readability.
• The guard logic could be extracted into a small helper to avoid repetition and improve testability.
• The test imports `JsonValue` solely for type annotation; the import could be omitted at runtime.
• Consider refactoring the finiteness check into a utility function like `isValidCount` and reuse it in `inferCredibility`.
• Add a test case where all three counts are non-finite to confirm the guard short-circuits early.

Signal	Result	Evidence
Code review	✅ No blockers	2 reviewers, synthesized
Linked issue	✅ No-issue rationale	PR body explains why no issue is linked.
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Review load	✅ 20/20	Readiness component derived from cached public PR metadata and labels; size label size:XS.
Validation evidence	✅ 25/25	PR body includes validation/test evidence.
Open PR queue	✅ 10/10	0 open PR(s), 0 likely reviewable.
Contributor context	✅ Confirmed Gittensor contributor	dripsmvcp; Gittensor profile; 1070 PR(s), 3 issue(s).
Gate result	✅ Passing	No configured blocker found.

Nits — 1 non-blocking

• No linked issue detected — If this PR is intended to solve an issue, link it explicitly in the PR body.

Review context

• Author: dripsmvcp
• Role context: outside_contributor
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: Rust, MDX, Python, TypeScript
• Official Gittensor activity: 1070 PR(s), 3 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Link the issue being solved, or explicitly explain why this is a no-issue PR.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Review load = cached public PR metadata such as size labels, changed paths, and preflight status.
• Open PR queue = repo-wide review pressure; it is not a PR quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

Review details

Generated from public PR metadata and the diff. Advisory only; deterministic signals remain authoritative.

The change expands the finiteness guard in `inferCredibility` to include stale and unlinked pull request counts, preventing NaN propagation and ensuring a fallback credibility of 0.8. A new test validates this behavior for malformed stale and unlinked values. All existing tests pass, confirming functional correctness.

Nits (5)

• The added comment in `preview.ts` is quite long and could be split for readability.
• The guard logic could be extracted into a small helper to avoid repetition and improve testability.
• The test imports `JsonValue` solely for type annotation; the import could be omitted at runtime.
• Consider refactoring the finiteness check into a utility function like `isValidCount` and reuse it in `inferCredibility`.
• Add a test case where all three counts are non-finite to confirm the guard short-circuits early.

_{🟩 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 Gittensory, a quiet PR intelligence layer for OSS maintainers.

• Re-run Gittensory review

[JSONbored]

MERGE. fix(scoring): guard credibility against non-finite stale/unlinked counts.

• Real bug: inferCredibility only finite-guarded merged; a non-numeric cached
  stale/unlinked count coerces to NaN and clamp(NaN,…)=NaN poisons credibilityObserved,
  credibilityMultiplier, estimatedMergedScore (preview.ts:331-333, 365-366).
• Fix correct & well-placed: guard broadened to all three counts before the arithmetic;
  Number.isFinite catches NaN+Infinity; fallback 0.8 == MIN_CREDIBILITY so the multiplier
  becomes 1 — no div-by-zero / downstream regression.
• Coverage complete & meaningful: every || arm tested both ways; asserts
  credibilityObserved===0.8 per malformed branch + toBeCloseTo(0.87,5) on the happy path.
• Already current with main; scope clean; no attribution.
  Note (non-blocking): commits are authored as "plind-junior" while the PR is @dripsmvcp.