fix(registration): align dogfood pack lane strategy with config recommendation

[galuis116]

Summary

buildSelfDogfoodRegistrationPack (src/services/self-dogfood-registration-pack.ts) derived directPrFirst from the registration readiness report's current-lane mode, but the same pack also returns the config recommendation. The two diverge, so the pack's contributorLaneStrategy text and directPrFirst flag could contradict the gittensorConfigRecommendation shown beside them. Closes #442.

// before
const issueDiscoveryReady = readiness.issueDiscoveryReadiness.ready && readiness.issueDiscoveryReadiness.recommendation === "enabled";
const directPrFirst = !issueDiscoveryReady && readiness.recommendedRegistrationMode !== "issue_discovery";

readiness.recommendedRegistrationMode reflects the repo's current lane (laneToMode(lane)), while the recommendation reflects the advised config (participationMode is direct_pr unless intake + config are excellent). So a repo currently registered for issue discovery but with degraded health gets recommendedRegistrationMode: "issue_discovery" -> directPrFirst: false -> "Issue-discovery intake is strong enough to keep a bounded issue-discovery lane", while the same pack recommends participationMode: "direct_pr". The inverse also happens (an excellent direct-PR repo recommended split reported directPrFirst: true).

Scope

• src/services/self-dogfood-registration-pack.ts -- derive directPrFirst = recommendation.recommended.participationMode === "direct_pr" so the lane strategy text and flag always match the recommendation in the same pack (removed the now-unused readiness-based computation).
• test/unit/self-dogfood-registration-pack.test.ts:
  • Corrected the excellent-config fixture test: its recommendation is split, so directPrFirst is now false (matching the recommendation) -- the old assertion true encoded the contradiction.
  • Added fail-on-revert: a currently-issue-discovery repo whose recommendation is direct_pr must report directPrFirst: true and direct-PR-first text (old code returned false).

Validation

• npx tsc --noEmit -- clean.
• npx vitest run for the dogfood pack + route suites -- 16/16.
• Full suite -- 1208 passed, 1 skipped (excluding the two unparseable upstream test files and one unrelated mcp-cli timeout that passes in isolation).
• Branch coverage 97.02% (above the 97% gate).

Safety

• No response-shape change; only the (now consistent) directPrFirst flag and contributorLaneStrategy text change, and only when the readiness current-lane mode diverged from the recommendation.
• The recommendation already drives the pack's config tradeoffs, so the lane strategy now agrees with the rest of the pack rather than contradicting it.

[ghost]

Note

Gittensory Gate skipped

PR closed before full evaluation. No late first comment was created.

Signal	Result	Evidence	Action
Gate result	⚠️ Skipped	#443 is no longer open.	No action.

---

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

[JSONbored]

@galuis116 this is good to land.

A few notes:

• directPrFirst now follows the recommendation shown in the same pack instead of mixing in current-lane readiness state.
• The tests cover both the excellent split-lane case and the degraded issue-discovery-to-direct-PR case.
• This resolves #442 cleanly.
• No code changes requested.