chore(deps): align knowledge runtime cohort

[drewstone]

Outcome

Aligns agent-knowledge 7.0.1 with agent-eval 0.135.4 and agent-interface 0.37.0 so consumers install one shared contract version.

Proof

• pnpm why reports exactly one agent-eval and one agent-interface version
• 518 tests passed; 12 skipped by their existing opt-in conditions
• lint, source and contract typechecks, build, packed-package verification passed
• official GEPA and SkillOpt integration verification exited successfully
• merge-tree against current origin/main succeeded

[tangletools]

✅ Auto-approved drewstone PR — 95baf719

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 · 2026-07-29T15:28:00Z}

[tangletools]

🟢 Value Audit — sound

Verdict	sound
Concerns	0 (none)
Heuristic	0.0s
Duplication	0.0s
Interrogation	29.4s (2 bridge agents)
Total	29.4s

💰 Value — sound

Bumps agent-knowledge to 7.0.1 and pins agent-eval 0.135.4 / agent-interface 0.37.0 so consumers install one shared contract stack; routine cohort alignment matching the repo's established pattern.

• What it does: Single-commit dependency alignment: package.json agent-eval 0.135.2->0.135.4 and agent-interface 0.36.0->0.37.0, version 7.0.0->7.0.1, plus matching pnpm-lock.yaml and a CHANGELOG 7.0.1 entry. No source code changes.
• Goals it achieves: Guarantee every consumer of @tangle-network/agent-knowledge 7.0.1 resolves exactly one agent-eval and one agent-interface version, eliminating older nested copies (trace-ingestion, source-identity, interface contracts). Same goal the repo pursued at 6.1.11.
• Assessment: Clean, minimal, in-grain. The patch is the exact tool this codebase uses for cohort alignment (see CHANGELOG 6.1.11 'Updated @tangle-network/agent-eval to 0.135.2 so knowledge improvement uses... without installing an older Eval copy'). Exact-version pins (no caret/range) on the @tangle-network deps are consistent with the existing style. CHANGELOG updated. No scope creep.
• Better / existing approach: none — this is the right approach. Searched git log on package.json (8 prior cohort/release commits all follow this exact pin+changelog pattern) and confirmed no other mechanism exists; the pnpm overrides block is intentionally reserved for transitive force-resolutions, not for first-party cohort pins, so using it here would fight the grain.
• 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

Clean patch bump that aligns agent-knowledge to one shared agent-eval 0.135.4 / agent-interface 0.37.0 contract version, with a fully consistent lockfile and no duplicate nested copies.

• Integration: Both deps are the foundational layer this package is built on. Grep found 100+ import sites across src/ and tests/ (e.g. src/optimization.ts:1, src/rag-optimization.ts:5-6, src/memory/holdout.ts:3-5, src/agent-candidate.ts:4) consuming canonicalJson, contentHash, RunRecord, CampaignStorage, AgentCandidateJsonValue, etc. Every consumer reaches the bumped versions transitively; nothing is gated or u
• Fit with existing patterns: Follows the repo's established cohort-alignment pattern exactly — the prior 'chore(release)' commits (1af43e6, 7398984, 746e0f0) and AGENTS.md's layering diagram (agent-knowledge imports agent-eval + agent-interface downward) make this the canonical way versions move together. The lockfile's transitive bump of agent-core 0.4.25→0.4.26 (which itself pins agent-interface@0.37.0) is precisely the 'on
• Real-world viability: Lockfile is internally consistent: exactly one agent-eval@0.135.4 and one agent-interface@0.37.0 entry (pnpm-lock.yaml lines 639/644/2762/2772), specifier matches resolved version, no nested duplicate copies that would reintroduce the older contract. CHANGELOG.md updated with a 7.0.1 entry. Survives a normal pnpm install.
• 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 · 20260729T152853Z}

[tangletools]

✅ No Blockers — 95baf719

Review health 100/100 · Reviewer score 92/100 · Confidence 75/100 · 1 finding (1 low)

glm: Correctness 92 · Security 92 · Testing 92 · Architecture 92

Reviewer score is advisory once the run is complete and the verdict has no blockers.

Full multi-shot audit completed 3/3 planned shots over 3 changed files. Global verifier still owns final merge decision.

🟡 LOW agent-interface minor bump (0.36 -> 0.37) under own patch bump (7.0.0 -> 7.0.1) — package.json

agent-interface moves a MINOR version while this package bumps only a PATCH. If any source file in this repo re-exports or widens a type from agent-interface, a downstream consumer of @tangle-network/agent-knowledge could see new surface without a matching minor bump here. Not a defect in package.json itself, but the global verifier should confirm the source shot (the .ts changes in this PR, or typecheck against the new cohort) shows no newly-exported shape from agent-interface surfacing transitively. Mitigation already in place: CHANGELOG 7.0.1 entry scopes this to 'cohort alignment', consistent with patch.

---

_{tangletools · 2026-07-29T15:30:22Z · trace}

[tangletools]

✅ Approved — 1 non-blocking finding — 95baf719

Full multi-shot audit completed 3/3 planned shots over 3 changed files. Global verifier still owns final merge decision.

Full immutable report for this review: trace

Summary comment for this run: full summary

---

_{tangletools · 2026-07-29T15:30:22Z · immutable trace}