Best possible solution:

Keep this bounded refactor open and land it after maintainer review/CI so Voice Call classic responses use the same sessionTarget identity seam as the embedded runner and realtime voice consult path.

Do we have a high-confidence way to reproduce the issue?

Not applicable; this PR is an internal refactor rather than a user-reported bug. The behavior claim is verifiable through the changed generateVoiceResponse call path, the embedded-runner contract, and the PR's direct runtime seam proof.

Is this the best way to solve the issue?

Yes; this is the narrowest maintainable fix because the embedded runner already owns sessionTarget resolution and the realtime Voice Call consult path already uses the same shape. Keeping the old sessionFile call in this migrated caller would preserve an internal compatibility path the repo policy is actively retiring.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 3848b9619fdb.

Label changes:

• add P3: This is a small internal session-runtime cleanup for one bundled plugin, with no reported user-facing regression or new product surface.
• add proof: sufficient: Contributor real behavior proof is sufficient. The PR body includes after-change terminal proof from a live local checkout showing the changed function passed sessionTarget, omitted sessionFile, and returned the expected response; focused tests and guard output support it.
• add rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🐚 platinum hermit and patch quality is 🐚 platinum hermit.
• add status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (terminal): The PR body includes after-change terminal proof from a live local checkout showing the changed function passed sessionTarget, omitted sessionFile, and returned the expected response; focused tests and guard output support it.

Label justifications:

• P3: This is a small internal session-runtime cleanup for one bundled plugin, with no reported user-facing regression or new product surface.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🐚 platinum hermit and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (terminal): The PR body includes after-change terminal proof from a live local checkout showing the changed function passed sessionTarget, omitted sessionFile, and returned the expected response; focused tests and guard output support it.
• proof: sufficient: Contributor real behavior proof is sufficient. The PR body includes after-change terminal proof from a live local checkout showing the changed function passed sessionTarget, omitted sessionFile, and returned the expected response; focused tests and guard output support it.

PR surface:

Source +1, Tests +55, Other +69. Total +125 across 4 files.

What I checked:

• Current main still uses legacy file identity: On current main, generateVoiceResponse resolves a transcript file with resolveSessionFilePath and passes sessionFile into runEmbeddedAgent, so the PR is not obsolete on main. (extensions/voice-call/src/response-generator.ts:294, 3848b9619fdb)
• PR uses the storage-neutral embedded-run target: At the PR head, the Voice Call caller passes sessionTarget with agentId, sessionId, sessionKey, and storePath, and no longer passes sessionFile directly. (extensions/voice-call/src/response-generator.ts:320, e797e86dbe88)
• Embedded runner contract supports this shape: resolveAgentRunSessionTarget accepts sessionTarget and resolves the current file-backed runtime target from agentId, sessionId, sessionKey, and storePath; runEmbeddedAgentInternal applies that identity before resolving the transcript target. (src/agents/run-session-target.ts:23, 3848b9619fdb)
• Embedded runner consumes resolved target before execution: runEmbeddedAgentInternal calls applyAgentRunSessionTargetIdentity, resolves the run session target, and backfills the runtime sessionFile internally for downstream file-backed internals. (src/agents/embedded-agent-runner/run.ts:625, 3848b9619fdb)
• Sibling Voice Call path already uses the same invariant: The realtime voice consult path passes the same sessionTarget shape into runEmbeddedAgent, which supports this PR as a one-sided classic-response migration rather than a new runtime pattern. (src/talk/agent-consult-runtime.ts:269, 3848b9619fdb)
• Boundary guard ratchets the migrated caller: The PR adds migratedEmbeddedAgentSessionTargetFiles and checks the Voice Call response generator for resolveSessionFilePath or sessionFile passed to runEmbeddedAgent. (scripts/check-session-accessor-boundary.mjs:430, e797e86dbe88)

Likely related people:

• jalehman: Recent merged Path 3/session-accessor work updated the same boundary guard and plugin runtime surfaces, and this PR continues that already-merged migration path rather than introducing a disconnected idea. (role: recent area contributor; confidence: high; commits: 7e2b2d2987b2, 4d4769c0d6a5, e797e86dbe88; files: scripts/check-session-accessor-boundary.mjs, test/scripts/check-session-accessor-boundary.test.ts, extensions/voice-call/src/response-generator.ts)
• Alix-007: Blame and file history show the current Voice Call response generator, embedded-run session target resolver, and sibling realtime consult path were introduced in commit 605aede.... (role: introduced behavior; confidence: medium; commits: 605aede38c10; files: extensions/voice-call/src/response-generator.ts, extensions/voice-call/src/response-generator.test.ts, src/agents/run-session-target.ts)

• 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
• 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
• 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
• 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
• 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
• 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
• 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.