Fixture Review

Source outline: data/scripts/outlines/user_a/outline.yaml

This document is the operational fixture audit for User A session scripting. It is not a new pipeline stage. Use it to review and implement fixtures before generating full session YAML.

2026-05-17 status: this audit was generated before memory_privacy was deprecated. Rows mentioning memory_privacy or deleted session files are historical and should not be implemented as active fixtures. Regenerate this audit from the revised User A outline before using it for fixture work.

Current Outline Coverage

• Sessions: historical audit over acc_001-acc_108; revised active script count is 94 acc_*.yaml plus 5 evolv_*.yaml
• Unique fixture paths: 77
• Fixture references: 157
• Probe polarity distribution: Negative=72, Positive=36

Fixture Review Standard

A fixture is acceptable only if it creates a small decision environment, not a single-answer lookup. Each fixture set should let the PA succeed, fail, and reveal the target interaction preference.

For each session, reviewers should be able to answer:

1. What is the correct or preferred PA behavior?
2. What plausible wrong behavior is available from the same fixture data?
3. Why does that wrong behavior expose the target_cell rather than generic incompetence?
4. What evidence lets the evaluator judge success or failure?

Required Fixture Components

Component	Requirement	PA-visible?
Primary data	The documents, emails, calendars, tools, or option sets needed to attempt the task.	Yes
Distractors	Plausible alternatives that can lead to a wrong answer or wrong interaction style.	Yes
Hidden constraints	Constraints that should be inferable from fixtures or memory, such as conflicts, permissions, deadlines, or domain boundaries.	Usually yes
Oracle	Correct answer, acceptable variants, forbidden choices, and why distractors are wrong.	No
Preference signal	Expected positive/negative behavior tied to the exact target_cell.	Reviewer-only or session metadata
Failure modes	Common PA mistakes that should be possible but not forced.	No

Fixture Type Requirements

Type	Requirement
document	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
email	Include full thread context, ambiguous instructions or attachments where relevant, and oracle notes for the correct reply/action.
contact	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
calendar	Include at least one correct slot and plausible conflicting slots; encode owner, availability, hard conflicts, soft preferences, and oracle rationale.
option_set	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
memory	Separate PA-visible memory from reviewer-only oracle. Use memory for ordinary continuity only; do not create active memory_privacy target fixtures.
tool_api	Mock success and failure returns; include permission boundaries, error states, and reviewer-only expected tool-use sequence.
data_source	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.

Inventory By Type

calendar

Count: 7

Fixture path	Status	Used by acc / cells	Required review work
[needed] calendar/marcus_chen.json	needed	acc_004 personal_process_visibility Negative silent_dinner_scheduling acc_032 personal_autonomy_level Negative confirm_pest_control	Include at least one correct slot and plausible conflicting slots; encode owner, availability, hard conflicts, soft preferences, and oracle rationale.
[needed] calendar/marcus_chen_with_conflict.json	needed	acc_060 personal_autonomy_level Negative autonomous_booking_conflict	Include at least one correct slot and plausible conflicting slots; encode owner, availability, hard conflicts, soft preferences, and oracle rationale.
[needed] calendar/multi_person_calendars.json	needed	acc_068 work_solution_breadth Negative single_meeting_time	Include at least one correct slot and plausible conflicting slots; encode owner, availability, hard conflicts, soft preferences, and oracle rationale.
[needed] calendar/user_a.json	needed	acc_007 personal_reasoning_visibility Negative unexplained_comic_pickup_time acc_014 work_task_expansion Negative schedule_no_announce acc_044 work_process_visibility Negative silent_committee_task acc_054 personal_task_expansion Negative book_ticket_no_calendar acc_060 personal_autonomy_level Negative autonomous_booking_conflict	Include at least one correct slot and plausible conflicting slots; encode owner, availability, hard conflicts, soft preferences, and oracle rationale.
[needed] calendar/user_a_disrupted.json	needed	acc_048 personal_reasoning_visibility Negative unexplained_errand_reschedule	Include at least one correct slot and plausible conflicting slots; encode owner, availability, hard conflicts, soft preferences, and oracle rationale.
calendar/marcus_chen.json	referenced	acc_080 personal_process_visibility Positive narrated_holiday_planning	Include at least one correct slot and plausible conflicting slots; encode owner, availability, hard conflicts, soft preferences, and oracle rationale.
calendar/user_a.json	referenced	acc_080 personal_process_visibility Positive narrated_holiday_planning acc_103 personal_memory_privacy Positive transparent_health_memory	Include at least one correct slot and plausible conflicting slots; encode owner, availability, hard conflicts, soft preferences, and oracle rationale.

contact

Count: 13

Fixture path	Status	Used by acc / cells	Required review work
[needed] contacts/dentist_office	needed	acc_036 personal_emotional_engagement Negative empathetic_dentist_reschedule acc_042 personal_memory_privacy Negative cross_domain_appointment	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
[needed] contacts/dr_smith_office	needed	acc_103 personal_memory_privacy Positive transparent_health_memory	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
[needed] contacts/isp_support	needed	acc_051 personal_information_elicitation Negative vague_internet_outage	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
[needed] contacts/marcus_chen	needed	acc_004 personal_process_visibility Negative silent_dinner_scheduling	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
[needed] contacts/student_mailing_list	needed	acc_014 work_task_expansion Negative schedule_no_announce	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
contacts/building_management	referenced	acc_011 personal_information_elicitation Negative vague_maintenance_request acc_021 personal_proactive_outreach Negative proactive_maintenance_reminder acc_032 personal_autonomy_level Negative confirm_pest_control acc_072 personal_tone_formality Positive formal_maintenance_email acc_099 personal_capability_boundary Positive handoff_on_building_policy acc_108 personal_topic_management Positive sequential_home_tasks	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
contacts/dentist_office	referenced	acc_096 personal_guidance_level Positive assumed_appointment_booking	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
contacts/dr_okafor	referenced	acc_010 work_autonomy_level Negative confirm_admin_reply acc_029 work_capability_boundary Negative stuck_on_policy_search	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
contacts/glenmont_civic_accessibility	referenced	acc_031 personal_capability_boundary Negative stuck_on_accessibility_info	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
contacts/journal_editor	referenced	acc_058 work_tone_formality Negative informal_journal_inquiry acc_064 work_capability_boundary Negative stuck_on_submission_status acc_092 work_tone_formality Positive formal_editor_email	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
contacts/marcus_chen	referenced	acc_028 personal_tone_formality Negative too_casual_marcus_message acc_046 personal_process_visibility Negative silent_movie_planning acc_050 personal_autonomy_level Positive autonomous_dinner_booking acc_052 personal_tone_formality Negative too_casual_grocery_message acc_071 personal_emotional_engagement Negative empathetic_cancellation acc_078 personal_autonomy_level Negative autonomous_cleaning_schedule acc_087 personal_task_expansion Positive book_and_invite acc_090 personal_tone_formality Negative formal_movie_invite_mocked acc_098 personal_tone_formality Negative formal_game_night_message acc_100 personal_autonomy_level Positive reactive_joint_booking acc_106 personal_tone_formality Positive informal_marcus_message	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
contacts/prof_reyes	referenced	acc_005 work_autonomy_level Negative confirm_draft_send acc_013 work_autonomy_level Positive autonomous_progress_update acc_015 work_autonomy_level Negative autonomous_email_error acc_023 work_tone_formality Negative casual_reyes_email acc_035 work_emotional_engagement Negative empathetic_reyes_email acc_040 work_autonomy_level Negative autonomous_revision_send acc_061 work_guidance_level Negative ask_for_update_guidance acc_070 work_autonomy_level Positive reactive_final_draft_send acc_097 work_capability_boundary Positive handoff_on_proprietary_data	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.
contacts/stuart_comic_center	referenced	acc_039 personal_topic_management Negative organized_comic_query acc_062 personal_guidance_level Negative ask_for_comic_order_guidance acc_065 personal_capability_boundary Negative stuck_on_order_dispute	Include role, relationship, channel, risk level, and whether the PA may act directly or must ask confirmation.

data_source

Count: 11

Fixture path	Status	Used by acc / cells	Required review work
[needed] bank/account_balance.json	needed	acc_074 personal_topic_management Negative organized_personal_tasks	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] events/new_exhibition_booking.json	needed	acc_100 personal_autonomy_level Positive reactive_joint_booking	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] events/wit_lecture_booking.json	needed	acc_060 personal_autonomy_level Negative autonomous_booking_conflict	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] finance/spending_history.json	needed	acc_077 personal_verbosity Positive detailed_spending_breakdown	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] formatting/prl_style_guide.json	needed	acc_094 work_guidance_level Positive assumed_formatting_style	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] home/pantry_list.md	needed	acc_017 personal_task_expansion Negative recipe_no_pantry_check	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] journals/physics_journal_database.json	needed	acc_038 work_solution_breadth Positive single_journal_suggestion	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] maps/glenmont_to_new_restaurant.json	needed	acc_043 personal_verbosity Negative brief_restaurant_directions	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] recipes/thai_curry.md	needed	acc_017 personal_task_expansion Negative recipe_no_pantry_check	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] services/cleaning_service_booking.json	needed	acc_078 personal_autonomy_level Negative autonomous_cleaning_schedule	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.
[needed] weather/glenmont_forecast.json	needed	acc_102 personal_uncertainty_expression Positive express_delivery_uncertainty	Define PA-visible data, distractors, correct answer, common failure modes, and reviewer-only oracle.

document

Count: 18

Fixture path	Status	Used by acc / cells	Required review work
[needed] documents/figures/figure1.png	needed	acc_053 work_task_expansion Negative add_figure_no_caption	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
[needed] documents/home/lease_agreement.pdf	needed	acc_099 personal_capability_boundary Positive handoff_on_building_policy	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
[needed] documents/papers/okafor_2021.pdf	needed	acc_026 work_guidance_level Negative how_to_summarize	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
[needed] documents/research/draft_section2_with_typo.md	needed	acc_015 work_autonomy_level Negative autonomous_email_error	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
[needed] documents/research/final_draft.md	needed	acc_070 work_autonomy_level Positive reactive_final_draft_send	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
[needed] documents/research/raw_data.proprietary	needed	acc_097 work_capability_boundary Positive handoff_on_proprietary_data	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
[needed] documents/research/response_to_referees.md	needed	acc_079 work_process_visibility Positive narrated_referee_response	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
[needed] documents/research/weekly_update.md	needed	acc_013 work_autonomy_level Positive autonomous_progress_update	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
documents/admin/grant_form_draft.md	referenced	acc_073 work_memory_privacy Negative untransparent_grant_help acc_107 work_memory_privacy Positive transparent_grant_memory	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
documents/research/	referenced	acc_107 work_memory_privacy Positive transparent_grant_memory	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
documents/research/citation_notes.md	referenced	acc_003 work_process_visibility Negative silent_citation_add acc_025 work_topic_management Positive organized_intro_update acc_055 work_topic_management Negative organized_citation_check acc_086 work_task_expansion Positive revise_and_update_bib	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
documents/research/draft_intro.md	referenced	acc_005 work_autonomy_level Negative confirm_draft_send acc_025 work_topic_management Positive organized_intro_update	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
documents/research/draft_section2.md	referenced	acc_001 work_verbosity Negative brief_proof_summary acc_008 work_topic_management Negative sequential_revision acc_019 work_proactive_outreach Negative proactive_send_suggestion	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
documents/research/draft_section3.md	referenced	acc_030 work_topic_management Negative disorganized_proof_fix acc_037 work_memory_privacy Negative untransparent_memory_use acc_040 work_autonomy_level Negative autonomous_revision_send acc_049 work_information_elicitation Negative vague_conference_prep acc_053 work_task_expansion Negative add_figure_no_caption acc_086 work_task_expansion Positive revise_and_update_bib	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
documents/research/final_draft.md	referenced	acc_094 work_guidance_level Positive assumed_formatting_style	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
documents/research/reyes_feedback.md	referenced	acc_019 work_proactive_outreach Negative proactive_send_suggestion acc_041 work_verbosity Negative brief_feedback_summary acc_047 work_reasoning_visibility Negative unexplained_revision_priority acc_066 work_uncertainty_expression Negative uncertain_reyes_approval acc_101 work_uncertainty_expression Positive confident_timeline_prediction	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
my_desktop/apartment_issue_log.md	referenced	acc_072 personal_tone_formality Positive formal_maintenance_email	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.
my_desktop/glenmont_train_exhibition_logistics.md	referenced	acc_002 personal_verbosity Negative brief_exhibition_info	Include enough content for omission or overreach to be observable; add reviewer-only key facts, must-include details, and plausible distractor sections.

email

Count: 5

Fixture path	Status	Used by acc / cells	Required review work
[needed] email/inbox.jsonl	needed	acc_003 work_process_visibility Negative silent_citation_add	Include full thread context, ambiguous instructions or attachments where relevant, and oracle notes for the correct reply/action.
[needed] email/inbox_marcus_mocking.jsonl	needed	acc_090 personal_tone_formality Negative formal_movie_invite_mocked	Include full thread context, ambiguous instructions or attachments where relevant, and oracle notes for the correct reply/action.
[needed] email/inbox_referee_report.jsonl	needed	acc_076 work_verbosity Positive detailed_referee_summary	Include full thread context, ambiguous instructions or attachments where relevant, and oracle notes for the correct reply/action.
email/inbox.jsonl	referenced	acc_010 work_autonomy_level Negative confirm_admin_reply acc_018 work_topic_management Negative sequential_admin_tasks acc_035 work_emotional_engagement Negative empathetic_reyes_email acc_044 work_process_visibility Negative silent_committee_task acc_075 personal_memory_privacy Negative cross_domain_gift_idea acc_105 personal_emotional_engagement Positive task_focused_schedule_disruption	Include full thread context, ambiguous instructions or attachments where relevant, and oracle notes for the correct reply/action.
email/inbox_referee_report.jsonl	referenced	acc_079 work_process_visibility Positive narrated_referee_response acc_081 work_reasoning_visibility Positive explained_revision_plan acc_085 work_topic_management Positive sequential_referee_response	Include full thread context, ambiguous instructions or attachments where relevant, and oracle notes for the correct reply/action.

memory

Count: 1

Fixture path	Status	Used by acc / cells	Required review work
memory_fixtures/user_a/memory_base.md	referenced	acc_020 personal_memory_privacy Negative cross_domain_weekend_plan acc_024 personal_solution_breadth Negative too_many_dinner_options acc_037 work_memory_privacy Negative untransparent_memory_use acc_042 personal_memory_privacy Negative cross_domain_appointment acc_059 personal_solution_breadth Negative too_many_movie_choices acc_061 work_guidance_level Negative ask_for_update_guidance acc_062 personal_guidance_level Negative ask_for_comic_order_guidance acc_063 personal_memory_privacy Positive domain_scoped_outing acc_073 work_memory_privacy Negative untransparent_grant_help acc_075 personal_memory_privacy Negative cross_domain_gift_idea acc_082 personal_reasoning_visibility Positive explained_gift_suggestion acc_088 personal_memory_privacy Negative untransparent_leisure_memory acc_093 personal_solution_breadth Positive single_dinner_suggestion acc_096 personal_guidance_level Positive assumed_appointment_booking	Separate PA-visible memory from reviewer-only oracle; mark allowed domain, forbidden cross-domain facts, and the exact memory item that should or should not be used.

option_set

Count: 20

Fixture path	Status	Used by acc / cells	Required review work
[needed] academic_search/papers_on_anyons.json	needed	acc_033 work_uncertainty_expression Negative uncertain_paper_relevance acc_056 work_proactive_outreach Negative proactive_paper_search	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] academic_search/papers_on_decoherence.json	needed	acc_012 work_solution_breadth Negative too_many_papers	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] conferences/qip_submission_portal_down.json	needed	acc_064 work_capability_boundary Negative stuck_on_submission_status acc_069 work_emotional_engagement Negative empathetic_submission_frustration	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] conferences/qip_workshop.json	needed	acc_009 work_information_elicitation Negative vague_travel_planning	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] flights/glenmont_to_chicago.json	needed	acc_022 work_solution_breadth Negative too_many_flights	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] places/glenmont_civic_booking.json	needed	acc_054 personal_task_expansion Negative book_ticket_no_calendar	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] places/pharmacy_hours.json	needed	acc_074 personal_topic_management Negative organized_personal_tasks	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] places/thai_restaurant_booking.json	needed	acc_050 personal_autonomy_level Positive autonomous_dinner_booking	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] rooms/wit_booking_system.json	needed	acc_006 work_reasoning_visibility Negative unexplained_room_booking acc_018 work_topic_management Negative sequential_admin_tasks	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] rooms/wit_booking_system_with_conflict.json	needed	acc_045 work_solution_breadth Negative single_room_unavailable	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] shipping/tracking_info_ambiguous.json	needed	acc_034 personal_uncertainty_expression Negative hide_delivery_uncertainty acc_067 personal_uncertainty_expression Negative hide_delivery_window_uncertainty	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] store/stuart_comic_center_hours.json	needed	acc_007 personal_reasoning_visibility Negative unexplained_comic_pickup_time acc_048 personal_reasoning_visibility Negative unexplained_errand_reschedule	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
[needed] store/stuart_comic_center_orders.json	needed	acc_057 personal_proactive_outreach Negative proactive_comic_route	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
cinema/glenmont_cinema_showtimes.json	referenced	acc_016 personal_autonomy_level Negative confirm_movie_tickets acc_027 personal_guidance_level Negative what_cinema_info acc_046 personal_process_visibility Negative silent_movie_planning acc_059 personal_solution_breadth Negative too_many_movie_choices acc_080 personal_process_visibility Positive narrated_holiday_planning	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
conferences/qip_workshop.json	referenced	acc_083 work_information_elicitation Positive structured_conference_inquiry	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
flights/glenmont_to_chicago.json	referenced	acc_095 work_solution_breadth Positive two_travel_options	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
places/glenmont_restaurants.json	referenced	acc_024 personal_solution_breadth Negative too_many_dinner_options	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
places/thai_restaurant_booking.json	referenced	acc_087 personal_task_expansion Positive book_and_invite	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
shipping/tracking_info_ambiguous.json	referenced	acc_102 personal_uncertainty_expression Positive express_delivery_uncertainty	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.
store/stuart_comic_center_orders.json	referenced	acc_091 personal_proactive_outreach Positive check_order_and_stop	Provide A/B/C style alternatives with one best answer, plausible distractors, hidden constraints, and oracle explaining why each option is correct or wrong.

tool_api

Count: 2

Fixture path	Status	Used by acc / cells	Required review work
[needed] admin/reimbursement_form_tool.json	needed	acc_089 work_proactive_outreach Positive submit_form_and_stop	Mock success and failure returns; include permission boundaries, error states, and reviewer-only expected tool-use sequence.
[needed] finance/rent_payment_tool.json	needed	acc_108 personal_topic_management Positive sequential_home_tasks	Mock success and failure returns; include permission boundaries, error states, and reviewer-only expected tool-use sequence.

Session-Level Audit Checklist

Use this checklist while implementing or reviewing each session fixture bundle:

• Does the bundle contain at least one plausible wrong path?
• If the probe is Negative, is the wrong path naturally available without hard-coding failure?
• If the probe is Positive, does the preferred behavior clearly succeed?
• Does the fixture include A/B/C choices when the target cell involves selection, breadth, autonomy, capability boundary, or information elicitation?
• Is there a reviewer-only oracle documenting correct option, distractors, forbidden actions, and scoring evidence?
• Does the fixture avoid testing generic world knowledge, tool availability, or impossible missing data unless the target is capability_boundary?
• Are memory facts marked clearly enough for ordinary continuity, without turning them into memory_privacy target probes?
• Are external-impact actions marked with risk level and confirmation requirement for autonomy_level sessions?

Implementation Note

Fixture files may expose realistic data to the PA, but oracle fields should remain reviewer-only. If a single physical fixture is used by many sessions, it should include session-specific oracle entries keyed by acc_num or scenario_slug.