wip(handoff): clean stale TODOs and plan issue cleanup

Co-Authored-By: Codex <noreply@openai.com>
This commit is contained in:
2026-05-01 01:47:41 -04:00
parent 595844de0b
commit a21fe93454
4 changed files with 92 additions and 12 deletions

View File

@@ -0,0 +1,74 @@
# Issue Cleanup Plan - 2026-05-01
## Tracker Hygiene
These are safe tracker updates before any feature work:
1. Close Gitea #127 (`feat: show AI content scope indicator`) unless an always-visible badge is still desired.
- Current code already has IT/MSP scope copy in the assistant empty state.
- `ASSISTANT_SYSTEM_PROMPT` also has an off-domain redirect boundary.
2. Rewrite Gitea #66 (`Tree Templates + Import/Export`) to the remaining scope only.
- `.rfflow` export/import is implemented in `tree_transfer.py` and exposed in the library UI.
- Remaining work: curated packs, authenticated one-click install from gallery, template versioning, marketplace/community path.
3. Close or archive open PR #124 (`feat/cockpit-harness`).
- It is unmergeable against current `main` and overlaps newer `/pilot` work.
4. Keep Gitea #58, #60, #128, #129, #130 open.
- They still describe real product gaps.
## Recommended Order
### 1. Low-Risk Maintenance
- Clean frontend lint warnings.
- Audit/remove stale eslint-disable comments.
- Add missing `data-testid` selectors for e2e-critical controls.
- Add observability around unexpected `currentChatRef` guard mismatches.
Why first: these reduce future regression cost and are small, well-bounded changes.
### 2. Pilot UX Friction
- #130: Add diagnostic command help affordances in `TaskLane` / action cards.
- #128: Decide whether task panel placement should be configurable or whether the existing responsive drawer is enough.
- EscalationQueue mobile design stays deferred until a customer asks for it.
Why second: this improves the current FlowPilot wedge without changing core data models.
### 3. Workflow Quality Signals
- #58: Add structured "step is wrong" flags separate from thumbs-up/down helpfulness.
- Existing `StepFeedback` is not enough; it only records helpful/unhelpful and cannot capture incorrect/outdated/unclear/missing-info reasons.
Why third: useful, but needs schema/API/UI/admin surfaces.
### 4. Client Intelligence
- #60: Recurring issue detection.
- Start with a read-only banner using existing `sessions.client_name + tree_id` filters.
- Add same-resolution detection only after confirming the available session outcome/node data is reliable enough.
Why fourth: high value, but it touches session-start and close-out flows and needs careful false-positive handling.
### 5. Documentation Structure
- #129: Hierarchical guide navigation.
- Current `/guides` route is a card grid plus detail pages with sections and breadcrumbs, but not a collapsible guide tree.
Why fifth: valid UX request, but less urgent than pilot workflow gaps.
## Gitea Actions Needed
The current environment does not have a Gitea token configured, so API writes fail with `401 token is required`. Once authenticated:
```bash
curl -X PATCH \
https://gitea.resolutionflow.com/api/v1/repos/chihlasm/resolutionflow/issues/127 \
-H "Authorization: token $GITEA_TOKEN" \
-H "Content-Type: application/json" \
-d '{"state":"closed"}'
```
For #66, prefer editing the title/body instead of closing it:
- Title: `feat: curated template packs and one-click install`
- Body: remove completed `.rfflow` export/import acceptance criteria and keep pack/install/versioning work.