chore(deps): align optimizer bridge versions

[drewstone]

Summary

• update Agent Knowledge to 5.0.1
• pin Agent Eval and the Python bridge to 0.126.6
• keep Runtime, Knowledge, and Bench on one optimizer provenance contract

Verification

• 466 tests passed, 12 skipped
• lint, typecheck, and build passed
• packed package clean install passed
• packed GEPA and SkillOpt integration passed 2/2 cases against public agent-eval-rpc 0.126.6
• merge-tree against current main passed

[tangletools]

✅ Auto-approved drewstone PR — 1d07c898

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 · 2026-07-24T21:06:48Z}

[tangletools]

🟢 Value Audit — sound

Verdict	sound
Concerns	0 (none)
Heuristic	0.0s
Duplication	0.0s
Interrogation	51.7s (2 bridge agents)
Total	51.7s

💰 Value — sound

Aligns the agent-eval/Python-bridge provenance contract from 0.126.5→0.126.6 and ships agent-knowledge 5.0.1; minimal, complete, in-grain.

• What it does: Bumps @tangle-network/agent-eval 0.126.5→0.126.6 in package.json, syncs the pnpm-lock.yaml, updates the README install snippet to the same version, adds a 5.0.1 CHANGELOG entry, and bumps the package's own version 5.0.0→5.0.1.
• Goals it achieves: Keep Knowledge, Runtime, and Bench on one shared optimizer provenance contract. The repo's verify-official-optimizers.mjs (scripts/verify-official-optimizers.mjs:17-34) enforces that the JS agent-eval pin and the Python agent-eval-rpc wheel are the same exact version; this PR moves that shared pin forward to 0.126.6 so all three consumers stay aligned.
• Assessment: Sound. The pin is exact (required by the regex gate at scripts/verify-official-optimizers.mjs:17), all four files move in lockstep, the lockfile is regenerated, and the CHANGELOG documents the rationale. The previous merged PR #79 established this contract at 0.126.5; this is the normal one-step follow-on. The PR body undercounts files (says 1, actually 4) but the code itself is complete and consi
• Better / existing approach: none — this is the right approach. The version has a single source of truth (package.json dependency), the verify script reads it rather than re-declaring it, and the README/CHANGELOG are documentation that is already in sync. No existing equivalent to reuse; the change IS the dependency declaration. Checked scripts/verify-official-optimizers.mjs and git log (0.126.5 in PR #79, 0.126.6 here — norm
• Model: opencode/zai-coding-plan/glm-5.2
• Bridge attempts: 2
• Bridge warning: opencode/kimi-for-coding/k2p7: bridge stream ended without value-audit content

🎯 Usefulness — sound

Aligns the deeply-consumed agent-eval substrate and its Python optimizer bridge to one pinned version (0.126.6), matching the established release-alignment pattern and enforced by the official-optimizer verification contract.

• Integration: Fully reachable and in active use: @tangle-network/agent-eval is the core substrate, imported across 77 sites (src/optimization.ts, src/memory/, src/benchmarks/, src/rag-, src/retrieval-, src/kb-improvement/*, src/release.ts, src/research-loop.ts). The version bump is not new surface — it satisfies a hard cross-language contract in scripts/verify-official-optimizers.mjs:16-34, which reads the
• Fit with existing patterns: Matches the established release discipline exactly. The prior 5.0.0 release did the identical operation for 0.126.5 (CHANGELOG.md:33), and the git log shows recurring 'chore(release): align agent stack' / 'chore(deps)' commits (e.g. 3cc4466). The bump is propagated consistently to package.json, pnpm-lock.yaml (resolved + integrity updated), README.md install line, and CHANGELOG.md. No competing pa
• Real-world viability: A patch-version bump of an exact-pinned core dependency, not new logic, so happy/error-path exposure is minimal and bounded by agent-eval's own semver. The alignment is guarded by the strict ^\d+...\d+$ pin check (verify-official-optimizers.mjs:17) and the cross-language version-equality throw (line 30), plus the packed integration test against the real published wheel — failure modes surface a
• Model: opencode/zai-coding-plan/glm-5.2
• Bridge attempts: 1

No concerns — sound change, no better or existing approach found. ✅

---

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 · 20260724T210933Z}