[hulk-linus]

Read the account's saved feedback preference with nansen feedback preferences, in human or JSON form. Agents use a fresh read to decide whether to ask before reporting. The submit command stays noninteractive; lasting choices belong in Superapp settings.

A feedback_not_enabled refusal stops automatic retries. Saved IDs and earlier uncertain outcomes remain available for recovery. No preference setter or admission check is added.

API-855. The shared MCP reader ships in the companion PR.

Validation: current-head CI passed the full suite on Node 22 and 24. Node 22 reports 4,952 passed and 2 skipped across 116 files. The final local feedback suites passed 85 tests; lint passed. All 8 controlled CLI host scenarios passed: enabled, off, unconfigured, lookup failure, approval, refusal, host denial and inability to ask. Actual outgoing HTTP calls were observed against a local upstream.

BI: existing command success/failure events remain accurate. Preference values and report content are not recorded; no event schema changes.

AI execution metadata

• Main writer/coordinator: OpenAI / Codex gpt-6.1-sol, high. Session 01a1489c-d3e4-7fb0-b2f2-2243876cef76.
• Combined main-task usage snapshot at 2026-10-17T07:54:21.422Z: 43,447,983 input, 42,562,432 cached input, 151,630 output including 80,360 reasoning, 43,599,613 total. This is shared implementation/orchestration usage across both API-855 PRs and cannot be partitioned. Later closeout overhead is excluded.
• Reviewer A: Factory / Anthropic claude-opus-5-5, high, verified from provider metadata. First pass session f9cb639c-d364-43cc-964b-35051a843e2a, input_tokens=40, cache_read_input_tokens=1478834, cache_creation_input_tokens=125319, output_tokens=22742, thinking_tokens=10833; total not supplied. Format correction session b1772f32-cb16-45f4-9cc6-b538155d155a, input_tokens=8, cache_read_input_tokens=37727, cache_creation_input_tokens=24694, output_tokens=4731, thinking_tokens=375; total not supplied. Distinct sessions; raw provider fields are preserved.
• Reviewer B: Cursor / xAI grok-4.7-high-fast, high and fast, verified from provider metadata. First pass session d40ada9b-9c72-4045-9b6d-f66ae8b8640c, inputTokens=145903, cacheReadTokens=3624960, cacheWriteTokens=0, outputTokens=47460; reasoning and total not supplied. Final verification: session 1402028a-5510-481d-ab0f-24474f8c144d, inputTokens=124182, cacheReadTokens=2174848, cacheWriteTokens=0, outputTokens=42677; reasoning and total not supplied. Final report format correction: session 0b7fc41f-5ef5-480d-b95e-e3dd06d1c75a, inputTokens=26831, cacheReadTokens=107904, cacheWriteTokens=0, outputTokens=4941; reasoning and total not supplied. Usage is separate from the main-task snapshot.
• Controlled reporting-host tests: OpenAI / Codex gpt-6.1-sol, high, 16 completed sessions shared across this PR set. Reported usage input_tokens=814864, cached_input_tokens=621312, cache_write_input_tokens=0, output_tokens=4651, reasoning_output_tokens=1621. Two initial timed-out starts supplied no completed-turn usage. These are policy tests, not writers or independent reviewers.

[nansen-pr-reviewer]

nansen-pr-reviewer summary

The change adds a fresh account-preference read and rollout-denial handling, with schema, documentation, changeset, and regression coverage. The earlier uncertainty bug for HTTP 5xx responses followed by rollout denial is fixed, but a final exhausted HTTP 5xx response on the explicit-ID submit path still loses the uncertain-outcome classification and can encourage an unsafe new submission.

Findings by Severity

Severity	Count
🟡 Medium	1

Deterministic check: success — Found 1 finding(s) within acceptable thresholds

Risk: 3/5 (Elevated)

✅ Auto-approved — Deterministic gates passed: severity thresholds met, recommendation "approve_with_comments", risk level 3/5 within threshold 3.

---

Token usage: 48,991 input, 1,662 output, 8,883 cache read | Usage Guide

Cooldown: for the next 10 minutes (counting from when this review finished), new pushes to this PR will not trigger another review — the next push after the window expires will. Need a fresh review sooner? Comment @nansen-pr-reviewer[bot] re-review.

[nansen-pr-reviewer]

✅ Auto-approved by Nansen review

Deterministic gates passed: severity thresholds met, recommendation "approve", risk level 2/5 within threshold 3.

• The deterministic severity check passed: No issues found
• Risk level 2/5 (Low)
• The model recommended: approve

Approval was decided by deterministic gates, not by the model. The model's recommendation can block auto-approval but never cause it.

[Codier]

Preserve receipt validation for feedback get and the earlier uncertainty warning after a rollout denial.

[nansen-pr-reviewer]

✅ Auto-approved by Nansen review

Deterministic gates passed: severity thresholds met, recommendation "approve_with_comments", risk level 3/5 within threshold 3.

• The deterministic severity check passed: Found 1 finding(s) within acceptable thresholds
• Risk level 3/5 (Elevated)
• The model recommended: approve_with_comments

Approval was decided by deterministic gates, not by the model. The model's recommendation can block auto-approval but never cause it.

[Codier]

No further code changes requested.

[nansen-pr-reviewer]

✅ Auto-approved by Nansen review

Deterministic gates passed: severity thresholds met, recommendation "approve_with_comments", risk level 3/5 within threshold 3.

• The deterministic severity check passed: Found 1 finding(s) within acceptable thresholds
• Risk level 3/5 (Elevated)
• The model recommended: approve_with_comments

Approval was decided by deterministic gates, not by the model. The model's recommendation can block auto-approval but never cause it.

[Codier]

Approved for merge.