chore(deps): move the first-party cohort onto agent-interface 1.0.0

[drewstone]

Problem

@tangle-network/traces@0.11.5 on npm pins a retired generation exactly: agent-eval 0.143.0, agent-runtime 0.126.0, sandbox 0.17.2. Each of those pins an agent-interface below 1.0.0 exactly.

Measured — install the first-party cohort at @latest in a clean directory and walk node_modules recursively:

@tangle-network/agent-interface: 7 physical copies
  1.0.0   node_modules/@tangle-network/agent-interface
  0.47.0  node_modules/@tangle-network/sdk-telemetry/node_modules/@tangle-network/agent-interface
  0.43.1  node_modules/@tangle-network/traces/node_modules/@tangle-network/agent-interface
  0.43.0  node_modules/@tangle-network/traces/node_modules/@tangle-network/agent-core/...
  0.43.0  node_modules/@tangle-network/traces/node_modules/@tangle-network/agent-eval/...
  0.43.0  node_modules/@tangle-network/traces/node_modules/@tangle-network/agent-knowledge/...
  0.43.0  node_modules/@tangle-network/traces/node_modules/@tangle-network/sandbox/...

Five of the seven come from this package.

main already carried newer pins (agent-eval 0.145.3, agent-runtime 0.133.2, sandbox 0.21.1), but under version 0.11.5 — the version the registry already holds. publish.yml publishes from a v* tag and npm refuses a republish, so those pins could never reach a consumer.

Change

The manifest moves to the cohort the rest of the fleet resolves: agent-eval 0.145.22, agent-runtime 0.137.0, sandbox 0.27.1. Each declares agent-interface as ^1.0.0, so one copy resolves.

@tangle-network/agent-interface 1.0.0 joins the manifest directly because agent-runtime 0.137.0 states it as a required peer.

The version moves to 0.11.6 so the change can publish.

Proof

No source changes were needed.

pnpm typecheck   exit 0
pnpm test        51 files, 658 tests, all passed
pnpm build       ESM 1663ms, DTS 12672ms, success

Scope

Manifest and lockfile. The release needs a v0.11.6 tag after merge.

[drewstone]

@tangletools review now

[tangletools]

✅ Auto-approved drewstone PR — 8ae10372

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.

This approval is provisional. It rests on the audit running. If the audit cannot run — for example the CLI bridge rejects it — this approval is dismissed rather than left standing, so an unrun check never reads as a passing one.

_{tangletools · auto-approval · reason: drewstone_author · 2026-08-16T21:18:48Z}

[tangletools]

✅ Auto-approved drewstone PR — 8ae10372

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.

This approval is provisional. It rests on the audit running. If the audit cannot run — for example the CLI bridge rejects it — this approval is dismissed rather than left standing, so an unrun check never reads as a passing one.

_{tangletools · auto-approval · reason: drewstone_author · 2026-08-16T21:18:53Z}

[tangletools]

🟢 Value Audit — sound

Verdict	sound
Coverage	1 of 2 lenses (usefulness)
Concerns	0 (none)
Heuristic	0.0s
Duplication	0.0s
Interrogation	523.3s (2 bridge agents)
Total	523.3s

⚠️ Partial audit — the verdict covers only usefulness. value: cli-bridge admission rejected (queue saturated). Treat the missing lens as unexamined, not as clear.

💰 Value — error

value agent never ran: the CLI bridge refused admission (no model was started).

• Model: opencode/deepseek/deepseek-v4-pro
• Bridge attempts: 4
• Bridge error: opencode/kimi-for-coding/k2p7: Bridge returned 503: bridge at capacity (queue_timeout, lane=reserved): active=20/20 queued=8/48 — no model was started

🎯 Usefulness — sound

A correct, verified dependency-cohort realignment that collapses 7 physical copies of agent-interface to 1, un-strands main's pins behind an already-published version, and lands through the repo's established exact-pin + tag-publish release path.

• Integration: Fully reachable. Every bumped dep is imported by shipped source: agent-eval across ~15 files (src/index.ts:138-163, src/report.ts:9, src/redact.ts:18, src/runtime-store.ts:15, src/replay-batch.ts:34, src/analyst-model-call.ts, src/otlp.ts:29, src/file-export.ts:19), agent-runtime via subpath import in src/supervisor-run-context.ts:45-46 (materializeTreeView + kernel types), sandbox's Sandbox class
• Fit with existing patterns: Matches the repo's established convention exactly: exact pins for the first-party cohort, caret for stable contracts (agent-trace-contract ^1.0.2), and the prior commit ba0bcb8 ('chore(deps): agent-eval 0.145.0...') shows this same bump-and-release pattern is the norm. Peer ranges cohere: agent-runtime 0.137.0 requires agent-eval >=0.145.21 <0.146.0 and the manifest pins 0.145.22; sandbox 0.27.1 i
• Real-world viability: Independently verified beyond the PR's claims: pnpm install --frozen-lockfile clean, pnpm typecheck exit 0, pnpm build ESM+DTS success, and the built CLI executed end-to-end against the new dependency graph (node dist/cli.js validate --otlp <fixture> produced a correct structured conformance report, exercising both ingest and error-reporting paths through agent-eval). The vitest suite coul
• Model: opencode/zai-coding-plan/glm-5.2
• Bridge attempts: 4

No concerns from the lens that ran (usefulness). The missing lens examined nothing, so this is not a full clean bill of health.

---

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 · 20260816T214715Z}