Bind Responses verification results to issued tool calls

[bigboateng]

Change contract

When: A Responses API client returns tool outputs to the local agent verification fixture.
Before: A session marker in a result for a missing or unrelated call ID can advance the allow/deny sequence. A result can also be accepted after a text-only response that never issued the call.
After: Results contribute evidence only when their call ID matches the issued phase call and that call was actually emitted. Valid conversation history remains accepted.

Boundaries preserved

• The allow and deny calls retain their existing IDs and session markers.
• Prior results and unrelated history do not invalidate a correctly correlated result.
• OpenCode retains its structured tool error checks for every matching result.
• Other model dialects and production Controller behavior remain unchanged.

Evidence

Scenario or check	Expected result	Observed result
Actual baseline HTTP sessions	Establish the verification gap and valid history behavior	37 independent sessions; 18 admission counterexamples retained
Candidate identity, phase, history and issuance cases	Reject unrelated or unissued results; accept valid results	37 independent HTTP cases and nine additional direct controls passed
Repository checks	Pass after source-bound native provenance refresh	51 targeted tests, Go, static checks and formatting passed; all 198 maintenance tests completed with one existing skip after authentic provenance refresh
Native captures	All supported agents and runtimes pass with current evidence	33 agent/platform cells and three runtime cells passed (capture run)
Exact-head PR CI/report	All required checks pass and report matches results	49 checks passed at ab9083b68a24; report confirms 33 real-agent and three runtime cells

Failure and recovery

Uncorrelated results cannot advance the fixture trajectory and receive its existing rejection response. No retries or alternate evidence paths are added. Revert the fixture change and refresh native provenance to recover the previous implementation.

Remaining limits

The counterexamples concern the verification fixture's evidence admission. They do not demonstrate a production Controller bypass. Existing marker matching and result semantics for other model dialects remain outside this change.

[github-actions]

Pitot coding-agent E2E

Intelligence Flow is the verification source; Pitot's public README carries the latest main status.

Agent	Ubuntu	macOS	Windows	Result	Evidence
Claude	✅ Pass · real CLI 2.1.296 · native · binary-observed anthropic messages	✅ Pass · real CLI 2.1.296 · native · binary-observed anthropic messages	✅ Pass · real CLI 2.1.296 · native · binary-observed anthropic messages	✅ Passing	All platforms are required
Cursor	✅ Pass · real CLI 2026.10.01-e373342 · native · binary-observed cursor connect proto	✅ Pass · real CLI 2026.10.01-e373342 · native · binary-observed cursor connect proto	✅ Pass · real CLI 2026.10.01-e373342 · native · binary-observed cursor connect proto	✅ Passing	All platforms are required
Codex	✅ Pass · real CLI 0.162.1 · native · binary-observed openai responses	✅ Pass · real CLI 0.162.1 · native · binary-observed openai responses	✅ Pass · real CLI 0.162.1 · native · binary-observed openai responses	✅ Passing	All platforms are required
Devin	✅ Pass · real CLI 3000.11.3 · native · binary-observed devin connect proto	✅ Pass · real CLI 3000.11.3 · native · binary-observed devin connect proto	✅ Pass · real CLI 3000.11.3 · native · binary-observed devin connect proto	✅ Passing	All platforms are required
DeepSeek Harness	✅ Pass · real CLI 0.2.0-rc.2 · native · binary-observed openai chat	✅ Pass · real CLI 0.2.0-rc.2 · native · binary-observed openai chat	✅ Pass · real CLI 0.2.0-rc.2 · native · binary-observed openai chat	✅ Passing	All platforms are required
GitHub Copilot CLI	✅ Pass · real CLI 1.0.95 · native · binary-observed openai chat	✅ Pass · real CLI 1.0.95 · native · binary-observed openai chat	✅ Pass · real CLI 1.0.95 · native · binary-observed openai chat	✅ Passing	All platforms are required
Gemini	✅ Pass · real CLI 0.63.0 · native · binary-observed gemini generate content	✅ Pass · real CLI 0.63.0 · native · binary-observed gemini generate content	✅ Pass · real CLI 0.63.0 · native · binary-observed gemini generate content	✅ Passing	All platforms are required
Kimi Code	✅ Pass · real CLI 2.1.1 · native · binary-observed openai chat	✅ Pass · real CLI 2.1.1 · native · binary-observed openai chat	✅ Pass · real CLI 2.1.1 · native · binary-observed openai chat	✅ Passing	All platforms are required
OpenCode	✅ Pass · real CLI 2.0.27 · native · binary-observed openai responses	✅ Pass · real CLI 2.0.27 · native · binary-observed openai responses	✅ Pass · real CLI 2.0.27 · native · binary-observed openai responses	✅ Passing	All platforms are required
Pi	✅ Pass · real CLI 1.1.0 · native · binary-observed openai chat	✅ Pass · real CLI 1.1.0 · native · binary-observed openai chat	✅ Pass · real CLI 1.1.0 · native · binary-observed openai chat	✅ Passing	All platforms are required
Qwen Code	✅ Pass · real CLI 0.25.1 · native · binary-observed openai chat	✅ Pass · real CLI 0.25.1 · native · binary-observed openai chat	✅ Pass · real CLI 0.25.1 · native · binary-observed openai chat	✅ Passing	All platforms are required

Runtime capabilities

Capability	Ubuntu	macOS	Windows	Result
pitot request	✅ Pass · real runtime allow/deny	✅ Pass · real runtime allow/deny	✅ Pass · real runtime allow/deny	✅ Passing

Source commit: ab9083b68a24