chore(release): agent-eval 0.100.3

[drewstone]

Summary

• bump agent-eval and agent-eval-rpc to 0.100.3
• publish the already-merged product-benchmark contract in the next tag/release
• add changelog entry for @tangle-network/agent-eval/product-benchmark

Checks

• pnpm typecheck
• pnpm test (256 files, 2633 passed, 2 skipped)
• python -m pip install -e .[dev] && pytest -q (18 passed)
• NODE_OPTIONS=--max-old-space-size=8192 pnpm build
• pnpm run verify:package
• git merge-tree --write-tree origin/main HEAD
• npm view @tangle-network/agent-eval@0.100.3 version returned 404 before publish

[tangletools]

✅ Auto-approved drewstone PR — 4b2fd5a9

This PR was opened by the trusted drewstone account.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: drewstone_author · 2026-07-01T15:09:51Z}

[tangletools]

🔴 Value Audit — redundant-or-flawed

Verdict	redundant-or-flawed
Concerns	2 (2 strong-concern)
Heuristic	0.0s
Duplication	0.0s
Interrogation	244.0s (2 bridge agents)
Total	244.0s

💰 Value — redundant-or-flawed

Bumps package versions to 0.100.3 consistently, but mishandles CHANGELOG.md by replacing the [Unreleased] section instead of moving its entries, causing a merge conflict with origin/main and dropping release notes for changes that will ship in the same release.

• What it does: Bumps @tangle-network/agent-eval and agent-eval-rpc from 0.100.2 to 0.100.3 in package.json:3, clients/python/pyproject.toml:7, and the PackageNotFoundError fallback in clients/python/src/agent_eval_rpc/__init__.py:61; adds a [0.100.3] entry to CHANGELOG.md:7 for the product-benchmark subpath.
• Goals it achieves: Publishes the already-merged product-benchmark subpath (package.json:117-120, tsup.config.ts:26) under a new tag and keeps the npm and PyPI package versions locked, correcting the 0.100.2 release where __init__.py was left behind and needed a follow-up fix (1aaa5f6).
• Assessment: The version bumps are correct and complete (all three version locations are now in sync). However, the CHANGELOG.md edit replaces the [Unreleased] section that existed on origin/main rather than moving those entries into [0.100.3]. As a result, release notes for the eval-fixture/campaign changes (47d2a4e) are lost even though that code is on main and will ship with 0.100.3. Additionally,
• Better / existing approach: Preserve the existing [Unreleased] entries by moving them under the new [0.100.3] section alongside the product-benchmark note, then resolve the merge with origin/main. This follows the Keep a Changelog convention already used in the file and ensures the release notes cover everything in the tag.
• Model: opencode/kimi-for-coding/k2p7
• Bridge attempts: 1

🎯 Usefulness — sound

A clean release-mechanics bump to 0.100.3 that publishes the already-merged product-benchmark subpath; versions are in sync across all three files and the new module is correctly wired into exports, build, and the verify script.

• Integration: Fully reachable. The product-benchmark subpath being published is wired into package.json exports (line 117), tsup build config (tsup.config.ts:26), and the verify-package-exports guard script (lines 37, 112). All three version literals (package.json:3, pyproject.toml:7, agent_eval_rpc/init.py:61) are synchronized at 0.100.3. Imminent caller per CHANGELOG: product agents consuming @tangle-netw
• Fit with existing patterns: Follows the established release pattern exactly — identical shape to the 0.100.2 (3ba4252) and 0.100.1 (223b7f8) releases, and the subpath-export + re-export-from-src/index.ts + verify-package-exports triplet is the same pattern used by ./perf, ./multishot, ./campaign, and every other substrate subpath. No competing or duplicate surface.
• Real-world viability: Version-literal bump plus CHANGELOG entry; no runtime logic changed. The init.py fallback literal at line 61 only fires when package metadata is absent (dev/egg install), which is the intended robustness behavior established by fix #294. No concurrency, error-path, or edge-input surface to evaluate.
• Model: opencode/zai-coding-plan/glm-5.2
• Bridge attempts: 1

💰 Value Audit

🔴 Branch has a CHANGELOG.md merge conflict with origin/main [against-grain] ``

git merge-tree --write-tree origin/main HEAD exits 1 with CONFLICT (content): Merge conflict in CHANGELOG.md. The release commit cannot land cleanly as-is. Evidence: merge-tree output shows three CHANGELOG.md versions (base, main, HEAD) and auto-merge fails.

🔴 CHANGELOG replaces [Unreleased] instead of moving its entries into [0.100.3] [against-grain] ``

origin/main has an [Unreleased] section documenting eval-fixture UX, planCampaignRun, dispatchRef, and a manifestHash resumability fix (47d2a4e, CHANGELOG.md:7-22 on main). The release commit removes all of those notes and replaces them with a single product-benchmark line (CHANGELOG.md:7-12 on HEAD). Since the code for those changes is on main and will be in the 0.100.3 tag, the release notes are now incomplete.

---

What this audit checks

It judges the change on its merits — not whether it was tasked out in an issue. Unticketed, fast-moving work is fine; the question is whether the change is good and whether a better or existing approach should be used instead.

Pass	What it asks
Heuristic	Vague title? Whitespace-only or cruft-bearing diff? (content signals only)
Duplication	Do added function/class names already exist elsewhere in the repo?
Value Audit	What does it do? What goal does it achieve? Is it good? Better architecture or already-exists?
Usefulness Audit	Does it integrate and fit? Will it hold up in real use and actually get used?

Findings are concerns, not blocks — the human reviewer decides what to do with them.

_{value-audit · 20260701T151654Z}

[tangletools]

✅ Auto-approved drewstone PR — 63d7e3e9

This PR was opened by the trusted drewstone account.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: drewstone_author · 2026-07-01T15:17:22Z}

[tangletools]

✅ Auto-approved drewstone PR — 95a61cb5

This PR was opened by the trusted drewstone account.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: drewstone_author · 2026-07-01T15:21:20Z}