@tangle-network/agent-bench 0.4.9 → 0.6.0

package/CHANGELOG.md

@@ -1,5 +1,16 @@
 # Changelog
 
+## 0.6.0
+
+- Consume Runtime 0.121.0 with Sandbox 0.16.0 so a clean install no longer requires the superseded Sandbox 0.15 line.
+- Release under a new minor because Sandbox is a pre-1.0 dependency moving across a minor boundary.
+
+## 0.5.0
+
+- Consume Runtime 0.120.0 with Eval 0.140.1, Interface 0.40.0, Knowledge 7.0.4, and Sandbox 0.15.2.
+- Released under a new minor, not a patch, because Knowledge crosses a major (6.1.11 -> 7.0.4) and Runtime crosses eleven minors. A `^0.4.9` range admits only patches, so shipping this alignment as 0.4.10 would hand every such consumer a major dependency move without a range change.
+- Resolve the supervisor tree from the `{kind:'event', root, event}` journal envelope, and take the optional `controlAdapter` on the DSPy RLM trace engine, both by way of Eval 0.140.1. Bench calls neither surface directly, so no adapter code changed.
+
 ## 0.4.9
 
 - Consume Runtime 0.109.2 with Eval 0.135.2 and Knowledge 6.1.11 so benchmark runs use the corrected paired promotion decisions.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-bench",
-  "version": "0.4.9",
+  "version": "0.6.0",
   "type": "module",
   "description": "Benchmark adapters and execution for agent-runtime across coding, tool-use, RAG, memory, browser, and terminal tasks.",
   "repository": {
@@ -25,11 +25,11 @@
     }
   },
   "dependencies": {
-    "@tangle-network/agent-eval": "0.135.2",
-    "@tangle-network/agent-interface": "0.36.0",
-    "@tangle-network/agent-knowledge": "6.1.11",
-    "@tangle-network/sandbox": "0.15.2",
-    "@tangle-network/agent-runtime": "0.109.2"
+    "@tangle-network/agent-eval": "0.140.1",
+    "@tangle-network/agent-interface": "0.40.0",
+    "@tangle-network/agent-knowledge": "7.0.4",
+    "@tangle-network/sandbox": "0.16.0",
+    "@tangle-network/agent-runtime": "0.121.0"
   },
   "devDependencies": {
     "@arethetypeswrong/cli": "0.18.5",

package/src/quant-arena/quant-loop.mts

@@ -88,11 +88,11 @@ export const PINNED_BASELINES: Record<string, GenerateSignals> = {
 /** The two demo author seats: the plain author and the quant lens. */
 export function defaultQuantProposers(): ProposerSpec[] {
   return [
-    { name: 'default-author', profile: 'default-author.profile.json', harness: 'claude' },
+    { name: 'default-author', profile: 'default-author.profile.json', harness: 'claude-code' },
     {
       name: 'quant-researcher',
       profile: join(QUANT_PROFILES_DIR, 'quant-researcher.profile.json'),
-      harness: 'claude',
+      harness: 'claude-code',
       lens:
         'Favor ONE economically-motivated effect (trend, mean reversion, vol targeting) with few parameters. ' +
         'State the regime in which it should work and keep turnover low enough that 15bps a side cannot eat the edge.',
@@ -218,7 +218,7 @@ async function claudeShot(opts: {
   const res = await run('claude', argv, {
     stdin: opts.prompt,
     cwd: opts.cwd,
-    env: proposerShotEnv('claude'),
+    env: proposerShotEnv('claude-code'),
     timeoutMs: opts.timeoutMs,
   })
   if (res.code !== 0) {

package/src/rollout-ledger/backfill-swe-arena.test.mts

@@ -158,7 +158,7 @@ async function buildFixtureTree(): Promise<void> {
       candidateIndex: 0,
       shot: 1,
       maxShots: 3,
-      harness: 'claude',
+      harness: 'claude-code',
       model: null,
       promptSha256: 'sha256:abc',
       startedAt: '2026-07-22T19:27:18.350Z',
@@ -256,7 +256,7 @@ describe('backfillSweArena', () => {
     expect(proposer?.messages).toHaveLength(2)
     expect(proposer?.outcome.reward).toBe(1)
     expect(proposer?.outcome.reward_source).toBe('swe-arena-official-judge/candidate-resolved-fraction')
-    expect(proposer?.policy).toMatchObject({ harness: 'claude', model: 'claude-fable-5' })
+    expect(proposer?.policy).toMatchObject({ harness: 'claude-code', model: 'claude-fable-5' })
     expect(proposer?.cost.tokens_out).toBe(462)
     expect(proposer?.task).toMatchObject({ suite: 'swe-arena-proposer', rep: 1 })
 

package/src/swe-arena/activation.test.mts

@@ -245,9 +245,9 @@ describe('activation-predicate prefilter', () => {
 
   it('kills a candidate without .improve/activation.json (stage activation-predicate) and passes one WITH it', async () => {
     const proposers: ProposerSpec[] = [
-      { name: 'with-predicate', harness: 'claude' },
-      { name: 'without-predicate', harness: 'claude' },
-      { name: 'invalid-predicate', harness: 'claude' },
+      { name: 'with-predicate', harness: 'claude-code' },
+      { name: 'without-predicate', harness: 'claude-code' },
+      { name: 'invalid-predicate', harness: 'claude-code' },
     ]
     const gen = fanOutLoopsGenerator(config(proposers), {
       author: async (proposer, args) => {
@@ -284,7 +284,7 @@ describe('activation-predicate prefilter', () => {
   })
 
   it('does not require a predicate when the gate is off (gen-4 behavior unchanged)', async () => {
-    const cfg = config([{ name: 'legacy', harness: 'claude' }])
+    const cfg = config([{ name: 'legacy', harness: 'claude-code' }])
     cfg.activationGate = false
     const gen = fanOutLoopsGenerator(cfg, {
       author: async (_p, args) => {

package/src/swe-arena/factory.test.mts

@@ -510,19 +510,19 @@ describe('judgeFactoryPatch', () => {
   it('gold (impl-only PR diff) resolves with full score', async () => {
     const { result } = await judgeFactoryPatch(goodInst, await goldImplPatch(goodInst))
     expect(result).toMatchObject({ resolved: true, score: 1, passed: 2, total: 2 })
-  })
+  }, 60_000)
 
   it('empty patch (bare base) fails: judge tests cannot even collect', async () => {
     const { result } = await judgeFactoryPatch(goodInst, '')
     expect(result).toMatchObject({ resolved: false, score: 0, passed: 0, total: 2 })
-  })
+  }, 60_000)
 
   it('partial impl earns partial credit, never resolved', async () => {
     // add correct, mul wrong → 1/2.
     const patch = await makePatch(goodInst, 'export const add = (a, b) => a + b\nexport const mul = (a, b) => a + b\n')
     const { result } = await judgeFactoryPatch(goodInst, patch)
     expect(result).toMatchObject({ resolved: false, score: 0.5, passed: 1, total: 2 })
-  })
+  }, 60_000)
 
   it('an unappliable patch is the candidate\'s failure (resolved false), not infra', async () => {
     const garbage = 'diff --git a/nope.txt b/nope.txt\n--- a/nope.txt\n+++ b/nope.txt\n@@ -1 +1 @@\n-x\n+y\n'

package/src/swe-arena/gepa-seat.test.mts

@@ -58,7 +58,7 @@ describe('validateGepaSeat', () => {
     expect(() => validateGepaSeat(seat())).not.toThrow()
     expect(() => validateGepaSeat(seat({ engine: 'omni', maxMetricCalls: 8 }))).not.toThrow()
     expect(isGepaSeat(seat())).toBe(true)
-    expect(isGepaSeat({ name: 'x', harness: 'claude' })).toBe(false)
+    expect(isGepaSeat({ name: 'x', harness: 'claude-code' })).toBe(false)
   })
 
   it('requires a surface inside the declared change-space', () => {
@@ -68,7 +68,7 @@ describe('validateGepaSeat', () => {
   })
 
   it('rejects harness-seat fields on an engine seat instead of silently ignoring them', () => {
-    expect(() => validateGepaSeat(seat({ harness: 'claude' }))).toThrow(/'harness' belongs to harness-authored/)
+    expect(() => validateGepaSeat(seat({ harness: 'claude-code' }))).toThrow(/'harness' belongs to harness-authored/)
     expect(() => validateGepaSeat(seat({ merge: true }))).toThrow(/'merge'/)
     expect(() => validateGepaSeat(seat({ model: 'x' }))).toThrow(/'model'/)
     expect(() => validateGepaSeat(seat({ profile: 'p.json' }))).toThrow(/'profile'/)
@@ -234,7 +234,7 @@ describe('captureProposerProvenance with a gepa seat', () => {
     return { code: 0, stdout: 'source', stderr: '' }
   }
   it('records engine, surface, gepa version, bridge module, and the python runtime as harnessVersion', async () => {
-    const record = await captureProposerProvenance([{ name: 'claude-author', harness: 'claude' }, seat()], {
+    const record = await captureProposerProvenance([{ name: 'claude-author', harness: 'claude-code' }, seat()], {
       exec: okExec,
       readSettingsModel: () => 'settings-model',
     })
@@ -251,7 +251,7 @@ describe('captureProposerProvenance with a gepa seat', () => {
     expect(gepa.harness).toBeUndefined()
     // The claude seat is untouched by the gepa capture path.
     expect(record.proposers.find((p) => p.name === 'claude-author')).toMatchObject({
-      harness: 'claude',
+      harness: 'claude-code',
       settingsModel: 'settings-model',
     })
   })

package/src/swe-arena/outer-loop.mts

@@ -646,7 +646,7 @@ export interface OuterLoopConfig {
    *  budget.maxImprovementShots; the LIB owns the dial (capabilities.mts
    *  fails loud on a substrate that would drop it). */
   maxShots: number
-  proposerHarness: 'claude' | 'codex' | 'opencode'
+  proposerHarness: 'claude-code' | 'codex' | 'opencode'
   proposerTimeoutMs: number
   /** GEN-3 proposer fan-out: N proposers author candidates CONCURRENTLY, each
    *  an AgentProfile-pinned harness invocation (see proposer-fanout.mts).
@@ -759,7 +759,7 @@ export function defaultRound4Config(
     // bootstrap run writes it; the lib validates it on every consumption.
     premeasuredBaselinePath: join(hh, 'r4', 'premeasured-baseline.json'),
     maxShots: 3,
-    proposerHarness: 'claude',
+    proposerHarness: 'claude-code',
     // Per author SHOT (agenticGenerator timeoutMs). 20 min timed out 3× under
     // degraded capacity in gen-1 ("author shot timed out") — doubled to 40 min.
     proposerTimeoutMs: 2_400_000,
@@ -858,18 +858,18 @@ export function defaultGen3Config(
   const base = defaultRound4Config(hh, opts)
   const outDirName = opts.outDirName ?? 'gen3'
   const proposers: ProposerSpec[] = [
-    { name: 'default-author', profile: 'default-author.profile.json', harness: 'claude' },
+    { name: 'default-author', profile: 'default-author.profile.json', harness: 'claude-code' },
     {
       name: 'mechanics-author',
       profile: 'default-author.profile.json',
-      harness: 'claude',
+      harness: 'claude-code',
       diagnosisSlice: 'mechanics',
       lens: 'Focus on MECHANICS: worker lifecycle, sandbox/clone contracts, settlement and delivery paths. Prefer code-path fixes over prompt wording.',
     },
     {
       name: 'prompts-author',
       profile: 'default-author.profile.json',
-      harness: 'claude',
+      harness: 'claude-code',
       diagnosisSlice: 'prompts',
       lens: 'Focus on PROMPTS: worker/brain instruction wording, placement guidance, self-check discipline. Prefer prompt/instruction changes over code-path rewrites.',
     },
@@ -945,10 +945,10 @@ export function defaultGen4Config(
 ): OuterLoopConfig {
   const base = defaultGen3Config(hh, { outDirName: opts.outDirName ?? 'gen4' })
   const proposers: ProposerSpec[] = [
-    { name: 'claude-author', profile: 'default-author.profile.json', harness: 'claude' },
+    { name: 'claude-author', profile: 'default-author.profile.json', harness: 'claude-code' },
     { name: 'glm-author', harness: 'opencode', model: 'zai-coding-plan/glm-5.2' },
     ...(opts.includeCodex === false ? [] : [{ name: 'codex-author', harness: 'codex' } satisfies ProposerSpec]),
-    { name: 'merge-author', profile: 'default-author.profile.json', harness: 'claude', merge: true },
+    { name: 'merge-author', profile: 'default-author.profile.json', harness: 'claude-code', merge: true },
   ]
   return {
     ...base,
@@ -1263,7 +1263,7 @@ const CODEX_AMBIENT_AUTH_VARS = ['OPENAI_API_KEY', 'OPENAI_BASE_URL'] as const
 
 export function proposerShotEnv(harness: OuterLoopConfig['proposerHarness']): NodeJS.ProcessEnv {
   const env: NodeJS.ProcessEnv = { ...process.env }
-  if (harness === 'claude') {
+  if (harness === 'claude-code') {
     for (const name of CLAUDE_AMBIENT_AUTH_VARS) delete env[name]
   }
   if (harness === 'codex') {

package/src/swe-arena/proc.test.mts

@@ -36,7 +36,7 @@ describe('run process-group timeout', () => {
       '-e',
       `const fs=require('node:fs'); const marker=process.argv[1]; process.on('SIGTERM',()=>{fs.writeFileSync(marker,'settled'); setTimeout(()=>process.exit(0),25)}); setInterval(()=>{},1000)`,
       marker,
-    ], { timeoutMs: 60, killGraceMs: 500 })
+    ], { timeoutMs: 1_000, killGraceMs: 500 })
 
     expect(result.code).toBe(TIMEOUT_RC)
     expect(result.timedOut).toBe(true)

package/src/swe-arena/proposer-fanout.mts

@@ -91,7 +91,7 @@ export interface ProposerSpec {
   profile?: string
   /** Required for harness-authored seats. Absent on an engine seat
    *  (`engine` set) — enforced both ways at generator construction. */
-  harness?: 'claude' | 'codex' | 'opencode'
+  harness?: 'claude-code' | 'codex' | 'opencode'
   /** GEN-4 pinned model id, threaded to the harness CLI as `-m <model>` via
    *  the author profile's `model.default` (harnessInvocation maps it for all
    *  three harnesses). Unset = the CLI's own resolved model (its login/settings
@@ -201,7 +201,7 @@ export function loadAuthorProfile(spec: ProposerSpec): AgentProfile | undefined
 
 /** The gen-2 author, codified: one bare-profile claude proposer. */
 export function defaultProposers(): ProposerSpec[] {
-  return [{ name: 'default-author', profile: 'default-author.profile.json', harness: 'claude' }]
+  return [{ name: 'default-author', profile: 'default-author.profile.json', harness: 'claude-code' }]
 }
 
 /** The profile the author shot actually runs: the loaded profile (if any) with

package/src/swe-arena/proposer-fanout.test.mts

@@ -73,7 +73,7 @@ describe('sliceFindings', () => {
 
 describe('proposerBuildPrompt', () => {
   it('appends the lens AFTER the shared protocol prompt, leaving the change-space text intact', () => {
-    const spec: ProposerSpec = { name: 'x', harness: 'claude', lens: 'Prefer code-path fixes.' }
+    const spec: ProposerSpec = { name: 'x', harness: 'claude-code', lens: 'Prefer code-path fixes.' }
     const prompt = proposerBuildPrompt({ findings: [] }, spec)
     expect(prompt).toContain('DECLARED CHANGE-SPACE')
     expect(prompt.indexOf('DECLARED CHANGE-SPACE')).toBeLessThan(prompt.indexOf('YOUR AUTHORING LENS (x)'))
@@ -81,7 +81,7 @@ describe('proposerBuildPrompt', () => {
   })
 
   it('is the bare round prompt without a lens', () => {
-    const spec: ProposerSpec = { name: 'x', harness: 'claude' }
+    const spec: ProposerSpec = { name: 'x', harness: 'claude-code' }
     expect(proposerBuildPrompt({ findings: [] }, spec)).not.toContain('AUTHORING LENS')
   })
 })
@@ -111,7 +111,7 @@ describe('resolveAuthorProfile (pinned models)', () => {
     const spec: ProposerSpec = {
       name: 'x',
       profile: 'default-author.profile.json',
-      harness: 'claude',
+      harness: 'claude-code',
       model: 'claude-fable-5',
     }
     const profile = resolveAuthorProfile(spec)
@@ -126,15 +126,15 @@ describe('resolveAuthorProfile (pinned models)', () => {
   })
 
   it('is byte-identical to loadAuthorProfile without a pin (gen-3 seats unchanged)', () => {
-    const spec: ProposerSpec = { name: 'x', profile: 'default-author.profile.json', harness: 'claude' }
+    const spec: ProposerSpec = { name: 'x', profile: 'default-author.profile.json', harness: 'claude-code' }
     expect(resolveAuthorProfile(spec)).toEqual(loadAuthorProfile(spec))
-    expect(resolveAuthorProfile({ name: 'bare', harness: 'claude' })).toBeUndefined()
+    expect(resolveAuthorProfile({ name: 'bare', harness: 'claude-code' })).toBeUndefined()
   })
 })
 
 describe('proposerBuildPrompt with pareto parents', () => {
   it('appends the parents section (evidence + diffs) after the protocol prompt and lens', () => {
-    const spec: ProposerSpec = { name: 'x', harness: 'claude', lens: 'Prefer code-path fixes.' }
+    const spec: ProposerSpec = { name: 'x', harness: 'claude-code', lens: 'Prefer code-path fixes.' }
     const prompt = proposerBuildPrompt({ findings: [] }, spec, PARENTS)
     expect(prompt).toContain('DECLARED CHANGE-SPACE')
     expect(prompt).toContain('PARETO PARENTS')
@@ -146,14 +146,14 @@ describe('proposerBuildPrompt with pareto parents', () => {
   })
 
   it('leaves the prompt untouched when no parents are seeded (gen-3 behavior)', () => {
-    const spec: ProposerSpec = { name: 'x', harness: 'claude' }
+    const spec: ProposerSpec = { name: 'x', harness: 'claude-code' }
     expect(proposerBuildPrompt({ findings: [] }, spec)).not.toContain('PARETO PARENTS')
     expect(parentsPromptSection(PARENTS)).toContain('measured evidence')
   })
 })
 
 describe('mergeAuthorPrompt', () => {
-  const spec: ProposerSpec = { name: 'merge-author', harness: 'claude', merge: true }
+  const spec: ProposerSpec = { name: 'merge-author', harness: 'claude-code', merge: true }
 
   it('keeps the change-space contract and presents BOTH parent diffs with the coherent-union task', () => {
     const prompt = proposerBuildPrompt({ findings: [] }, spec, PARENTS)
@@ -182,11 +182,11 @@ describe('defaultGen4Config', () => {
     expect(config.proposers).toHaveLength(4)
     expect(config.populationSize).toBe(4)
     const byName = Object.fromEntries(config.proposers!.map((p) => [p.name, p]))
-    expect(byName['claude-author']).toMatchObject({ harness: 'claude', profile: 'default-author.profile.json' })
+    expect(byName['claude-author']).toMatchObject({ harness: 'claude-code', profile: 'default-author.profile.json' })
     expect(byName['claude-author']!.model).toBeUndefined()
     expect(byName['glm-author']).toMatchObject({ harness: 'opencode', model: 'zai-coding-plan/glm-5.2' })
     expect(byName['codex-author']).toMatchObject({ harness: 'codex' })
-    expect(byName['merge-author']).toMatchObject({ harness: 'claude', merge: true })
+    expect(byName['merge-author']).toMatchObject({ harness: 'claude-code', merge: true })
   })
 
   it('drops the codex seat (and shrinks the population) when includeCodex is false', () => {
@@ -216,7 +216,7 @@ describe('defaultGen4Config', () => {
 
 describe('loadAuthorProfile', () => {
   it('loads the committed default-author profile (bare: no prompt, no model)', () => {
-    const profile = loadAuthorProfile({ name: 'a', profile: 'default-author.profile.json', harness: 'claude' })
+    const profile = loadAuthorProfile({ name: 'a', profile: 'default-author.profile.json', harness: 'claude-code' })
     expect(profile?.name).toBe('swe-arena-default-author')
     expect(profile?.prompt).toBeUndefined()
     expect(profile?.model).toBeUndefined()
@@ -224,7 +224,7 @@ describe('loadAuthorProfile', () => {
   })
 
   it('returns undefined without a profile path', () => {
-    expect(loadAuthorProfile({ name: 'a', harness: 'claude' })).toBeUndefined()
+    expect(loadAuthorProfile({ name: 'a', harness: 'claude-code' })).toBeUndefined()
   })
 
   it('fails loud when a non-codex proposer declares profile resources (they would be dropped)', async () => {
@@ -232,7 +232,7 @@ describe('loadAuthorProfile', () => {
     try {
       const path = join(dir, 'with-resources.json')
       await writeFile(path, JSON.stringify({ name: 'r', resources: { files: [] } }))
-      expect(() => loadAuthorProfile({ name: 'a', profile: path, harness: 'claude' })).toThrow(/silently drop/)
+      expect(() => loadAuthorProfile({ name: 'a', profile: path, harness: 'claude-code' })).toThrow(/silently drop/)
     } finally {
       await rm(dir, { recursive: true, force: true })
     }
@@ -302,7 +302,7 @@ describe('defaultGen3Config', () => {
 
   it('defaultProposers codifies the gen-2 author: one bare-profile claude entry', () => {
     expect(defaultProposers()).toEqual([
-      { name: 'default-author', profile: 'default-author.profile.json', harness: 'claude' },
+      { name: 'default-author', profile: 'default-author.profile.json', harness: 'claude-code' },
     ])
   })
 })
@@ -357,19 +357,25 @@ describe('fanOutLoopsGenerator', () => {
 
   it('authors ALL proposers concurrently in separate worktrees and applies each patch to its candidate slot', async () => {
     const proposers: ProposerSpec[] = [
-      { name: 'alpha', harness: 'claude' },
-      { name: 'beta', harness: 'claude' },
+      { name: 'alpha', harness: 'claude-code' },
+      { name: 'beta', harness: 'claude-code' },
     ]
     let inFlight = 0
     let maxInFlight = 0
     const seenWorktrees: string[] = []
+    let releaseAuthors!: () => void
+    const bothAuthorsStarted = new Promise<void>((resolve) => {
+      releaseAuthors = resolve
+    })
     const gen = fanOutLoopsGenerator(baseConfig(proposers), {
       author: async (proposer, args) => {
         inFlight++
         maxInFlight = Math.max(maxInFlight, inFlight)
         seenWorktrees.push(args.worktreePath)
-        // Hold the slot across a tick so real parallelism is observable.
-        await new Promise((r) => setTimeout(r, 30))
+        if (inFlight === proposers.length) releaseAuthors()
+        // Neither author may finish before both have started. A serial implementation
+        // therefore fails this test instead of passing or failing with machine speed.
+        await bothAuthorsStarted
         // Inside the declared change-space (extensions/pi/**) so the always-on
         // change-space pre-filter does not kill the candidate.
         await mkdir(join(args.worktreePath, 'extensions', 'pi'), { recursive: true })
@@ -390,7 +396,7 @@ describe('fanOutLoopsGenerator', () => {
 
     expect(first).toMatchObject({ applied: true, label: 'alpha' })
     expect(second).toMatchObject({ applied: true, label: 'beta' })
-    expect(first.rationale).toContain('proposer alpha (claude)')
+    expect(first.rationale).toContain('proposer alpha (claude-code)')
     // Both patches landed on the SAME driver worktree across the two calls —
     // the driver discards/finalizes between calls in production; here we only
     // assert each call applied its own proposer's file.
@@ -401,8 +407,8 @@ describe('fanOutLoopsGenerator', () => {
 
   it('kills a candidate at the smoke pre-filter: applied=false, no patch applied, kill recorded with reason', async () => {
     const proposers: ProposerSpec[] = [
-      { name: 'good', harness: 'claude' },
-      { name: 'bad', harness: 'claude' },
+      { name: 'good', harness: 'claude-code' },
+      { name: 'bad', harness: 'claude-code' },
     ]
     const config = baseConfig(proposers)
     config.prefilter = { enabled: true, smokeInstance: 'cheapest-of-set' }
@@ -455,7 +461,7 @@ describe('fanOutLoopsGenerator', () => {
   })
 
   it('kills an out-of-space diff at the change-space pre-filter before any smoke spend', async () => {
-    const proposers: ProposerSpec[] = [{ name: 'rogue', harness: 'claude' }]
+    const proposers: ProposerSpec[] = [{ name: 'rogue', harness: 'claude-code' }]
     const config = baseConfig(proposers)
     config.prefilter = { enabled: true, smokeInstance: 'cheapest-of-set' }
     let smokeRan = false
@@ -482,7 +488,7 @@ describe('fanOutLoopsGenerator', () => {
   })
 
   it('returns applied:false without a kill when a proposer authors nothing', async () => {
-    const gen = fanOutLoopsGenerator(baseConfig([{ name: 'idle', harness: 'claude' }]), {
+    const gen = fanOutLoopsGenerator(baseConfig([{ name: 'idle', harness: 'claude-code' }]), {
       author: async () => ({ applied: false, summary: '' }),
     })
     expect((await gen.generate(generatorArgs(0))).applied).toBe(false)
@@ -490,7 +496,7 @@ describe('fanOutLoopsGenerator', () => {
   })
 
   it('fails loud when candidateIndex exceeds the proposer list (populationSize drift)', async () => {
-    const gen = fanOutLoopsGenerator(baseConfig([{ name: 'only', harness: 'claude' }]), {
+    const gen = fanOutLoopsGenerator(baseConfig([{ name: 'only', harness: 'claude-code' }]), {
       author: async () => ({ applied: false, summary: '' }),
     })
     await expect(gen.generate(generatorArgs(1))).rejects.toThrow(/populationSize must equal/)
@@ -514,7 +520,7 @@ describe('fanOutLoopsGenerator', () => {
   })
 
   it('refuses a merge seat without >=2 materialized parents', () => {
-    const config = baseConfig([{ name: 'merge-author', harness: 'claude', merge: true }])
+    const config = baseConfig([{ name: 'merge-author', harness: 'claude-code', merge: true }])
     expect(() => fanOutLoopsGenerator(config, { author: async () => ({ applied: false, summary: '' }) })).toThrow(
       /merge proposer/,
     )
@@ -533,8 +539,8 @@ describe('fanOutLoopsGenerator', () => {
   it('rejects duplicate proposer names and empty proposer lists', () => {
     expect(() =>
       fanOutLoopsGenerator(baseConfig([
-        { name: 'dup', harness: 'claude' },
-        { name: 'dup', harness: 'claude' },
+        { name: 'dup', harness: 'claude-code' },
+        { name: 'dup', harness: 'claude-code' },
       ])),
     ).toThrow(/duplicate/)
     expect(() => fanOutLoopsGenerator({ ...defaultRound4Config(), loopsRepo, outDir })).toThrow(/empty/)
@@ -585,7 +591,7 @@ describe('gen-5 prompt sections', () => {
   }
 
   it('appends EVIDENCE MAP + briefing + activation contract after the protocol prompt', () => {
-    const spec: ProposerSpec = { name: 'x', harness: 'claude' }
+    const spec: ProposerSpec = { name: 'x', harness: 'claude-code' }
     const prompt = proposerBuildPrompt({ findings: [] }, spec, [], {
       briefing,
       activationGate: true,
@@ -598,7 +604,7 @@ describe('gen-5 prompt sections', () => {
   })
 
   it('the merge seat gets the gen-5 sections too', () => {
-    const spec: ProposerSpec = { name: 'merge-author', harness: 'claude', merge: true }
+    const spec: ProposerSpec = { name: 'merge-author', harness: 'claude-code', merge: true }
     const prompt = proposerBuildPrompt({ findings: [] }, spec, PARENTS, {
       briefing,
       activationGate: true,
@@ -609,7 +615,7 @@ describe('gen-5 prompt sections', () => {
   })
 
   it('leaves gen-3/gen-4 prompts byte-identical when no extras are passed', () => {
-    const spec: ProposerSpec = { name: 'x', harness: 'claude' }
+    const spec: ProposerSpec = { name: 'x', harness: 'claude-code' }
     const legacy = proposerBuildPrompt({ findings: [] }, spec, PARENTS)
     expect(legacy).not.toContain('EVIDENCE MAP')
     expect(legacy).not.toContain('ACTIVATION PREDICATE')

package/src/swe-arena/proposer-provenance.mts

@@ -24,6 +24,7 @@
 import { readFileSync } from 'node:fs'
 import { homedir } from 'node:os'
 import { join } from 'node:path'
+import { localHarnessExecutable } from '@tangle-network/agent-runtime/mcp'
 import {
   DEFAULT_GEPA_PYTHON,
   isGepaSeat,
@@ -117,10 +118,13 @@ export async function captureProposerProvenance(
     (h): h is NonNullable<ProposerSpec['harness']> => h !== undefined,
   )
   for (const harness of harnesses) {
-    const res = await exec(harness, ['--version'])
+    // The harness id is not the binary name (`claude-code` runs `claude`); read the executable
+    // from the runtime's harness table rather than spawning the id.
+    const executable = localHarnessExecutable(harness)
+    const res = await exec(executable, ['--version'])
     if (res.code !== 0) {
       throw new Error(
-        `proposer provenance: '${harness} --version' failed (rc=${res.code}) — the ${harness} seat cannot author. ` +
+        `proposer provenance: '${executable} --version' failed (rc=${res.code}) — the ${harness} seat cannot author. ` +
           `stderr: ${res.stderr.slice(0, 300)}`,
       )
     }
@@ -163,7 +167,7 @@ export async function captureProposerProvenance(
         harness: spec.harness,
         pinnedModel: spec.model ?? null,
         harnessVersion: versionByHarness.get(spec.harness!)!,
-        settingsModel: spec.harness === 'claude' && !spec.model ? readSettingsModel() : null,
+        settingsModel: spec.harness === 'claude-code' && !spec.model ? readSettingsModel() : null,
         authStatus: (spec.harness !== undefined ? authByHarness.get(spec.harness) : undefined) ?? null,
         merge: spec.merge === true,
       }

package/src/swe-arena/proposer-provenance.test.mts

@@ -9,10 +9,10 @@ import type { ProposerSpec } from './proposer-fanout.mts'
 const ok = (stdout: string) => ({ code: 0, stdout, stderr: '' })
 
 const gen4ish: ProposerSpec[] = [
-  { name: 'claude-author', profile: 'default-author.profile.json', harness: 'claude' },
+  { name: 'claude-author', profile: 'default-author.profile.json', harness: 'claude-code' },
   { name: 'glm-author', harness: 'opencode', model: 'zai-coding-plan/glm-5.2' },
   { name: 'codex-author', harness: 'codex' },
-  { name: 'merge-author', profile: 'default-author.profile.json', harness: 'claude', merge: true },
+  { name: 'merge-author', profile: 'default-author.profile.json', harness: 'claude-code', merge: true },
 ]
 
 describe('captureProposerProvenance', () => {
@@ -31,7 +31,7 @@ describe('captureProposerProvenance', () => {
     const byName = Object.fromEntries(record.proposers.map((p) => [p.name, p]))
     // Claude seat: no pin — the CLI's resolved settings model is the record.
     expect(byName['claude-author']).toMatchObject({
-      harness: 'claude',
+      harness: 'claude-code',
       pinnedModel: null,
       settingsModel: 'claude-fable-5',
       harnessVersion: 'claude-version 9.9.9',

package/src/swe-code-improve.mts

@@ -10,7 +10,7 @@
  * Wiring (all verified in this worktree):
  *   - improve()/codeProposerFor + rawTraceContext come from the LOCAL agent-runtime build, linked into
  *     this bench's node_modules (bench/node_modules/@tangle-network/agent-runtime -> /home/drew/code/agent-runtime).
- *   - The candidate proposer is agenticGenerator(harness:'claude'), BUT the shipped runLocalHarness
+ *   - The candidate proposer is agenticGenerator(harness:'claude-code'), BUT the shipped runLocalHarness
  *     spawns `claude --headless -p` and --headless is an unknown option on the current CLI (exit 1, no
  *     edits ever). We pass code.generator with a corrected runHarness that spawns
  *     `claude -p <prompt> --dangerously-skip-permissions` so the coding agent can actually edit the
@@ -286,7 +286,7 @@ async function main(): Promise<void> {
   }
 
   const generator = agenticGenerator({
-    harness: 'claude',
+    harness: 'claude-code',
     verify,
     timeoutMs: harnessTimeoutMs,
     // eslint-disable-next-line @typescript-eslint/no-explicit-any