---
id: RREG-WP-0004
type: workplan
title: "Repository Ability Registry - Characteristic Classification And Navigation"
domain: capabilities
repo: repo-registry
status: active
owner: codex
topic_slug: foerster-capabilities
created: "2026-04-29"
updated: "2026-04-29"
state_hub_workstream_id: "ad67787f-a89d-4cde-957c-18ef39b43912"
---

## Goal

Make repository profiles easier to understand and refine by adding first-class
classification and navigation support for characteristics. The product should
help curators move from repository scope to abilities, capabilities, features,
support, and observed facts while preserving flexibility for overlap and messy
real-world abstraction.

## P0: Characteristic Classification Fields

```task
id: RREG-WP-0004-T01
status: done
priority: high
state_hub_task_id: "fd2664c2-eb33-42f3-9624-c74bb0d30456"
```

Add `primary_class` and `attributes` to approved and candidate abilities,
capabilities, and features. Keep the migration additive, preserve existing
feature `type` compatibility, and default existing data to useful classes.

Acceptance: API/UI serialization includes classification fields for approved and
candidate graph elements; existing repositories migrate without data loss; tests
cover default backfill and candidate-to-approved promotion.

Implementation note 2026-04-29: approved and candidate abilities, capabilities,
and features now carry `primary_class` and `attributes`. The migration is
additive, existing features default their primary class and first attribute from
legacy `type`, API response models expose the new fields, and approval preserves
candidate classification metadata.

## P1: Classification Review UI

```task
id: RREG-WP-0004-T02
status: done
priority: high
state_hub_task_id: "ec9a2676-9e41-4f2d-8436-80670e5e051f"
```

Expose primary class and attributes in manual add/edit forms and candidate review
actions. Let curators filter characteristic lists by primary class and secondary
attributes without losing approved/candidate status filters.

Acceptance: a curator can edit the class and attributes of an approved or
candidate ability, capability, or feature and immediately use those values for
filtering.

Implementation note 2026-04-29: manual add/edit forms and element-list edit
forms expose primary class and attributes for approved characteristics. Candidate
abilities, capabilities, and features can also be edited with classification
metadata, and element listings can filter by primary class and secondary
attribute.

## P1: Deterministic Classification Proposals

```task
id: RREG-WP-0004-T03
status: done
priority: medium
state_hub_task_id: "d1f9646e-ac74-4fad-9183-7d5c4d5c93e0"
```

Teach deterministic candidate generation to propose conservative primary classes
and attributes from scanner facts, file surfaces, routes, manifests, tests,
documentation, and provider/config signals.

Acceptance: repo-registry and llm-connect analyses produce useful
classification labels while LLM assistance remains optional and disabled paths
remain smooth.

Implementation note 2026-04-29: deterministic candidate generation now proposes
conservative classification metadata. Abilities can classify as
`repository-intelligence`, `ai-integration`, or `developer-tooling`; capabilities
include `interface`, `llm-integration`, and `repository-structure`; features add
surface/provider attributes such as `api`, `cli`, `http`, `llm-provider`,
`openrouter`, `claude`, `credential`, and `fallback-policy`.

## P1: Characteristic Drilldown Navigation

```task
id: RREG-WP-0004-T04
status: todo
priority: medium
state_hub_task_id: "14ede41f-a0cb-4a9a-b4ba-f23b34d7ae33"
```

Improve navigation from high-level scope and ability views down to capabilities,
features, support, and observed facts. Facts should feel like drilldown evidence,
not the primary orientation surface.

Acceptance: the repository page and element listings make it natural to move
from scope to abilities to lower-level support/facts, with counts, filters, and
clear breadcrumbs.

## P2: Classification Quality Feedback

```task
id: RREG-WP-0004-T05
status: todo
priority: low
state_hub_task_id: "691f3cb7-c8a2-4f80-a6c2-29cb5a0c7a96"
```

Capture classification expectation gaps and review smells, including missing
primary classes, wrong feature surface classes, overbroad attributes, and
same-level/upward support patterns that indicate suboptimal organization.

Acceptance: reviewers can record classification-specific improvement inputs that
feed the scanner coevolution workflow.