feat(miner): leakage-safe task generation for historical replay

[dhgoal]

Summary

Implements leakage-safe task generation for the historical-replay calibration harness (Closes #3011).

New module packages/gittensory-miner/lib/replay-task-generation.js selects calibration-worthy freeze points, scrubs forward references out of the frozen context, tags each point's recency pool, and returns the frozen snapshot and the revealed post-T ground truth as separate bundles. Every function is pure and deterministic — no clock, no randomness, no IO — so a given (candidate, context) always yields an identical task.

• selectFreezePoint(candidate, thresholds) — a freeze point is eligible only with enough real history on both sides of T (minPriorCommits / minRevealedCommits); each unmet threshold is reported as a reason.
• detectForwardReferences / scrubForwardReferences(text, context) — classify and remove references that resolve only to post-T state:
  • scrubbable (self-delimited, replaced with a fixed placeholder): #N and GitHub issues/pull deep-links with N > knownIssueMax, and raw/URL commit SHAs not in the pre-T knownCommitShas set;
  • unscrubbable (detected but left in place): a bare integer that exactly matches a known post-T issue number — a bare number can't be blanket-removed without destroying legitimate pre-T numbers, so its presence must fail the freeze point rather than be silently mangled. (A plain decimal is never misread as a SHA — SHAs must carry a hex letter.)
• lintFrozenContext(texts, context) — the check that rejects a freeze-point candidate if any text still carries an unscrubbable forward reference after scrubbing.
• classifyRecencyPool(candidate, { modelCutoffIso }) — records the recent/older pool provenance (ISO-8601 sorts lexicographically, so no clock is needed) for later pool-sliced scoring.
• generateReplayTask(candidate, context, options) — one-shot: select → lint → scrub, returning frozen and revealed as separate top-level bundles (never merged, so the replay pipeline never holds both at once); ineligible or un-scrubbable candidates are rejected without producing a task.

Pure, self-contained, additive — a new foundation module alongside replay-objective-anchor. No existing module, CLI surface, or dependency changes; no src/** touched.

Scope

• Conventional Commit title; focused; linked issue feat(miner): leakage-safe task generation from real historical outcomes #3011; new lib/*.js + hand-written .d.ts (package convention), registered in the build node-check line. No site//CNAME/**/lovable/**; no CHANGELOG.md.

Validation

• npm run build --workspace @jsonbored/gittensory-miner (node --check, incl. the new module)
• npm run typecheck
• git diff --check; command-reference:check + docs:drift-check clean
• 17 unit tests: forward-ref detection (scrubbable vs unscrubbable, pre-T kept, decimal-not-a-SHA), scrubbing + placeholder + residual, lint pass/fail, freeze-point selection thresholds, recency-pool split + defaults, and generateReplayTask eligible (frozen/revealed separation asserted) / selection-reject / unscrubbable-reject / determinism.

Safety

• Deterministic — no clock, no randomness, no network, no IO. No secrets/wallets/hotkeys/trust/reward terms; the module scrubs GitHub issue/PR/commit references and applies threshold selection, exposing no internal scoring, reward, or trust state.

[superagent-security]

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

[loopover-orb]

Warning

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏸️ Gittensory review result - manual review recommended

_{Review updated: 2026-07-04 16:47:27 UTC}

4 files · 1 AI reviewer · no blockers · readiness 80/100 · CI green · clean

⏸️ Suggested Action - Manual Review

Review summary
The new replay task generation module is coherent: it selects freeze points, separates frozen/revealed bundles, scrubs self-delimited forward references, and rejects residual bare post-T issue numbers. The tests exercise the main pure paths and the package build script now syntax-checks the new module. I do not see a reachable correctness defect in the provided diff, but there are a few edge-case hardening points worth tightening before this becomes a production calibration primitive.

Nits — 5 non-blocking

• nit: packages/gittensory-miner/lib/replay-task-generation.js:90 treats any unknown 7-40 hex token with a letter as a commit SHA, so ordinary hex identifiers in frozen prose can be over-redacted unless the caller only feeds commit-shaped fields.
• nit: packages/gittensory-miner/lib/replay-task-generation.js:149 accepts negative threshold values because it only checks Number.isInteger, which makes a malformed threshold silently easier to satisfy instead of degrading to 0 or rejecting the candidate.
• nit: packages/gittensory-miner/lib/replay-task-generation.js:167 relies on raw lexicographic comparison of whatever strings the caller provides; that is fine for normalized ISO-8601 UTC strings, but mixed offsets or date-only strings can classify unexpectedly.
• packages/gittensory-miner/lib/replay-task-generation.js:90: consider requiring stronger SHA context for raw-token scrubbing, or document that the input text fields are known to contain commit prose where unknown hex tokens should be treated as leakage.
• packages/gittensory-miner/lib/replay-task-generation.js:149: clamp threshold defaults with `Number.isInteger(value) && value >= 0 ? value : 0` to match the candidate count handling and avoid malformed negative thresholds.

Signal	Result	Evidence
Code review	✅ No blockers	1 reviewer
Linked issue	✅ Linked	#3011
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	❌ 5/25	Preflight is holding this PR: the review lane is unavailable, so it is not ready for automated review.
Contributor workload	✅ 10/10	Author activity: 51 registered-repo PR(s), 30 merged, 0 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	dhgoal; Gittensor profile; 51 PR(s), 0 issue(s).
Gate result	✅ Passing	No configured blocker found.

Review context

• Author: dhgoal
• Role context: outside_contributor
• Public audience mode: oss maintainer
• Lane context: Repository registration is not available in the local Gittensory cache.
• Public profile languages: not available
• Official Gittensor activity: 51 PR(s), 0 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Await review-lane availability.
• Refresh registry data or choose a registered active repo.

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.

_{🟩 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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.09%. Comparing base (66262ce) to head (d667723).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3146   +/-   ##
=======================================
  Coverage   96.09%   96.09%           
=======================================
  Files         265      265           
  Lines       29088    29088           
  Branches    10595    10595           
=======================================
  Hits        27951    27951           
  Misses        493      493           
  Partials      644      644           

🚀 New features to boost your workflow:

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

[loopover-orb]

Gittensory approves — the gate is satisfied and CI is green.