test(tui): CLI ↔ TUI infuse parity + subrole sibling listing (proposal 003 PR-6 — closes proposal 003)

[flg77]

Summary

PR-6 of proposal 003 (ACC TUI usability hardening). Final PR of the series — closes proposal 003. After this merges, all six PRs (#54, #55, #56, #57, #58, #59) are on main and we can cut v0.2.0.

Two deliverables

1. CLI ↔ TUI infuse parity test

tests/test_infuse_parity.py (new, 7 cases). Pins what's actually equivalent between the CLI's _cmd_infuse and the TUI's action_apply ROLE_UPDATE payloads. Catches regression in either direction.

Test	Pin
test_envelope_byte_for_byte_parity	Top-level fields (signal_type, agent_id, collective_id, signature, approver_id) match byte-for-byte after stripping ts
test_both_paths_use_same_signal_type	signal_type=ROLE_UPDATE on both
test_both_paths_carry_role_definition_key	Top-level role_definition key present on both
test_role_definition_intersection_carries_same_values	Recursive intersection-only diff — values match for the 9 fields the TUI emits (incl. nested category_b_overrides)
test_documents_tui_field_omissions	TUI keys MUST be subset of CLI keys; logs the current omission set so operator can see the gap
test_no_secrets_in_either_payload	Neither path leaks api_key= / password= / private_key= / secret= / token= strings
test_both_paths_publish_on_same_subject	subject_role_update(cid) == "acc.{cid}.role_update" — rename catches

Known structural gap surfaced by the test (deferred): the CLI emits the full pydantic model_dump() (~15+ fields); the TUI form emits a 9-field subset. Closing this — either extending the TUI form or teaching the arbiter to default-fill — is a follow-up.

2. Ecosystem subrole sibling listing

acc/tui/screens/ecosystem.py gains _subrole_siblings() + _format_subrole_section() helpers. When the selected role has sibling directories matching <role>_* with a role.yaml, they're listed under a "Subroles (directory-derived)" Markdown section appended to the role.md view.

coding_agent → lists coding_agent_architect, coding_agent_dependency, coding_agent_implementer, coding_agent_reviewer, coding_agent_tester.

The section is explicitly labelled directory-derived because the first-class parent_role field is deferred to proposal 004.

Tests

94/94 green across the full TUI test surface (8 test files):

• test_infuse_parity.py (7 cases, new)
• test_ecosystem_screen_pilot.py (46, +8 new subrole cases)
• test_coding_agent_tui_pilot.py (8)
• test_configuration_screen_pilot.py (11)
• test_perf_soma_context_pilot.py (8)
• test_prompt_screen_pilot.py (13)
• test_file_picker_pilot.py (3)
• test_oversight_tui_diagnose.py (1)

CHANGELOG

[Unreleased] gains the subrole listing entry, a new Tests subsection documenting the parity test, and a "Known parity gap (deferred follow-up)" subsection that makes the structural divergence operator-visible.

Closing proposal 003

When this PR merges:

PR	State
#54 (PR-1)	✅ merged
#55 (PR-2)	✅ merged
#56 (PR-3)	✅ merged
#57 (PR-4)	✅ merged
#58 (PR-5)	✅ merged
#59 (PR-6)	closes the series

Cut v0.2.0 after merge:

git tag -a v0.2.0 -m "v0.2.0 — TUI usability hardening (proposal 003)"
git push --tags

Reference

Proposal 003 — …\ACC Implementation\003 - ACC TUI usability hardening.md (operator vault).

🤖 Generated with Claude Code