[sandl99]

Failure

The Rust integration tests in crates/nemoclaw-e2e/tests/integration/rust/ repeated checks that other tests own. Two pairs tested the same failure, and several tests compared outputs or exports that SDK or CLI unit tests already cover. This is section 3 of the duplicated-checks list on #14060.

Part of #14060.

Decision

Test	Change	Owner of what was removed
deployment::startup_and_observation_failures… + sandbox_startup_failure_names_reason…	Merged into a_sandbox_startup_failure_is_reported_safely_retained_and_recovered: the CLI's JSON report and plan, the unfinished-apply guard, the retry, recovery, failed and cut-off observations, and a lost delete	One test owns them now
deployment::failed_first_configuration… + fabric_deployment::failed_configuration…	Merged into fabric_deployment::a_rejected_start_is_reported_safely_and_a_corrected_document_replaces_the_sandbox	One test owns them now
export_observations::mixed_search_export…	Dropped	SDK unit export_preserves_search_definitions_and_rejects_registration_drift, export_requires_shared_definition_registrations_to_agree_on_credentials, and the drift table in authored_settings_survive_cli_export…
deployment::scoped_provider_labels…	Dropped	SDK inline_and_scoped_provider_sources_preserve_authored_names_without_values and CLI resource_results_name_authored_sources_and_separate_deferred_work
export_observations::export_refreshes…	Now export_skips_gateway_readiness_and_changes_nothing	The failed read moved into the startup failure test
export_observations::credentialed_inference…	Now a_credentialed_deployment_exports_and_tears_down_without_its_secret	Unchanged reapply: HCL scenarios. Observed reference export: SDK unit export_updates_only_the_observed_provider_definition. No probes: kubernetes-external-three-agents
deployment::interrupted_create…	Trimmed to the lost create, the pending-target guard, and an unrelated change finishing it	Exact counts, the CLI destroy, and re-apply after destroy were incidental
deployment::failed_first_apply…	One row, the agent start failure	The startup failure test owns the failed sandbox
deployment::cli_terminal_outputs…	Keeps the real-bundle lifecycle and JSON contract	CLI formatter tests own the wording
deployment::cli_redaction…	Now cli_redaction_withholds_short_and_colliding_credentials, without the recovery tail	The Fabric start failure test

Kept unchanged after checking:

• fabric_deployment::missing_runtime_declaration…: the SDK's unfinished_apply_without_complete_runtime_state_explains_how_to_recover checks a different case, a record whose apply is unfinished.
• deployment::rejected_policy…: the SDK's sandbox_admission tests check planning order with a scripted OpenTofu, not a gateway's admission refusal.

Rust integration tests: 35 to 31. Net −471 lines.

Validation

• cargo ci on darwin_arm64: all steps passed. Lifecycle ran 96 tests, down from 100 on v1.
• The 14 tests in deployment, export_observations, fabric_deployment, and multi_agent pass against a bundle built from this branch.

[sandl99]

Authoring session — opening

Field	Status	Value	Evidence or limit
Harness	partial	Pi coding agent 1.2.0	The session record names the Pi session format; the version comes from the active install path, not the record
Skills	recorded	nemoclaw-contributor-create-pr, v1 copy (complete read)	Skill reads to cutoff; other skills were listed only
Instruction files	recorded	AGENTS.md of main (injected); AGENTS.md of v1 (complete read); WRITING.md (complete read); .github/PULL_REQUEST_TEMPLATE.md (complete read); .github/SESSION_REPORT_TEMPLATE.md (complete read)	Startup injection and reads to cutoff
Tool calls	partial	typescript: 563	Wrapper calls, including failed calls, to cutoff; nested shell, file, and HTTP calls inside the wrapper are not counted
Session window	recorded	2026-11-02T05:58Z to 2026-11-03T11:26Z; 29 h 28 min	First event to cutoff before publication, including idle time; the session also authored #14063, #14121, and #14142
Compactions	recorded	0	Complete record to cutoff; context elisions are not compactions
Models	recorded	claude-opus-5-5	Model change event and assistant turns; no changes
Reasoning levels	recorded	medium; high	Thinking-level changes, in order

[github-actions]

v1 documentation preview: https://nvidia-preview-nemoclaw-v1-pr-14145.docs.buildwithfern.com/nemoclaw