[juliusmarminge]

Problem

In the SwiftUI app, tapping a completed delegated task in the Agents sheet fails with thread_not_found. Delegated thread IDs contain literal percent escapes; the Swift client escapes the path a second time, so the host looks up a different ID.

Change

Build environment request URLs with an explicitly percent-encoded path helper. Keep the plain-path helper for discovery, pairing, and route probes. Add a regression assertion on the outgoing bounded snapshot URL for a delegated thread ID containing %3A.

Scope and approval

Fixes #17643

Triaged bug via the automated issue flow. Targets t3code/rebuild-mobile-app-swift, which owns the native SwiftUI client.

Verification

• Real main server at 8d3b061121, isolated synthetic thread projection: the single-encoded bounded request returned HTTP 200 with the original thread ID; the double-encoded request returned HTTP 404 thread_not_found. Runtime transcript.
• Added V2OrchestrationClientTests.testThreadReadEncodesLiteralPercentInThreadIDOnce in the existing transport test file. Swift tests were not executed: Linux has no Swift compiler or Xcode. No claim of an observed red or green XCTest result.
• git diff --check passed. Two independent reviews (gpt-6.1-sol and claude-opus-5-5) returned MERGE with no confirmed code findings.
• No UI rendering code changes. Fixture matrix and evidence index explicitly list the unverified Swift/native rows. Native before/after captures, recordings, visual diffs, and “what else changed” are unavailable; the host transcript does not substitute for them.

Evidence	Before	After	Diff0	DiffT
Native SwiftUI interaction and static rows	Not captured (4619c0bca6)	Not captured (f8dadbc5f6)	Unavailable	Unavailable

cannot-verify ios-swift: native build, focused XCTest execution, and opening the delegated child on an Apple runtime remain for Mac verification.

Models: gpt-6.1-sol and claude-opus-5-5. Harnesses: Codex and Claude Agent via T3 Code.

[macroscopeapp]

Approvability

Verdict: Approved at f8dadbc

Macroscope's review found this PR approvable — This is a narrowly scoped iOS URL-encoding fix that preserves percent-encoded delegated-task thread IDs and adds a focused regression test. Existing static API routes and query behavior remain unchanged.

^{You can add or adjust custom eligibility rules. Learn more.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ No successful main baseline artifact is available yet. This run establishes the initial measurement.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	—	5.0 KiB	—	6.8 KiB	✅
Codex	Thread snapshot wire	—	3.8 KiB	—	4.9 KiB	✅
Codex	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Codex	Live turn WebSocket decoded	—	20.9 KiB	—	29.3 KiB	✅
Codex	Live turn messages	—	2	—	8	✅
Claude	Total thread wire	—	5.0 KiB	—	6.8 KiB	✅
Claude	Thread snapshot wire	—	3.8 KiB	—	4.9 KiB	✅
Claude	Live turn WebSocket wire	—	1.2 KiB	—	2.0 KiB	✅
Claude	Live turn WebSocket decoded	—	21.2 KiB	—	29.3 KiB	✅
Claude	Live turn messages	—	2	—	8	✅

Baseline: unavailable · PR result: f8dadbc · Source CI: failure

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.