---
id: capability.agents.custodian-runtime-tooling
name: Custodian Runtime And Repo-Classification Tooling
summary: 'Generic tooling from The Custodian''s runtime and ecosystem-management surface: an agent runtime
  framework (context, actions, tool adapters, policies) and repo-classification batch tooling used across
  the workstation''s 61 repos.'
owner: the-custodian
status: draft
domain: custodian
tags:
- agents
- runtime
- classification
maturity:
  discovery:
    current: D2
    target: D4
    confidence: medium
    rationale: This repo's README and canon content are marked confidential/proprietary; this entry deliberately
      scopes to the generic, non-proprietary tooling surface only (runtime/ agent framework, tools/ classification
      scripts), which is separately reusable and does not disclose canon/memory contents.
  availability:
    current: A1
    target: A2
    confidence: medium
    rationale: '`runtime/agent.py` plus context/actions/tool_adapters/policies modules form a real Python
      package (own pyproject, tests, README) under `runtime/`; `tools/` holds standalone scripts (batch_author_repo_classifications.py,
      human_review_classifications.py, validate_repo_classification.py, push_repo_classifications.sh,
      patch-forgejo-remote-urls.sh) used for ecosystem-wide repo classification and the Gitea-to-Forgejo
      remote URL migration.'
external_evidence:
  completeness:
    level: C1
    confidence: low
    basis: scope_vs_intent_and_consumer_expectations
    satisfied_expectations:
    - standalone agent-runtime package under runtime/ with its own tests
    - repo-classification batch/review/validate/push scripts used across the workstation roster
    - a patch-forgejo-remote-urls.sh script directly relevant to REUSE-WP-0019
    broken_expectations: []
    out_of_scope_expectations: []
  reliability:
    level: R0
    confidence: low
    basis: consumer_quality_signals
    known_reliability_risks:
    - repo root is confidential/proprietary (NDA); this entry's scope is deliberately narrow and must
      not be widened to canon/memory contents without explicit review
discovery:
  intent: Expose only the generic, non-confidential tooling that emerged from The Custodian's ecosystem-management
    work — an agent runtime framework and repo-classification batch tooling — as reusable, without describing
    canon or memory content.
  includes:
  - runtime/ agent framework (context, actions, tool adapters, policies)
  - repo-classification batch/review/validate/push scripts
  - patch-forgejo-remote-urls.sh
  excludes:
  - canon/ (curated knowledge substrate) and memory/ (operational logs) — confidential, out of registry
    scope entirely
  - governance/constitution content
  assumptions: []
  use_cases: []
  research_memos: []
availability:
  current_level: A1
  target_level: A2
  current_artifacts:
  - '`runtime/` Python package'
  - '`tools/*.py` and `tools/*.sh` classification scripts'
  target_artifacts: []
  consumption_modes:
  - library import (runtime/)
  - cli (tools/ scripts)
relations:
  depends_on: []
  supports: []
  related_to: []
evidence:
  documentation:
  - runtime/README.md
  tests:
  - runtime/tests/
  consumer_feedback: []
  bug_reports: []
  incidents: []
consumer_guidance:
  recommended_for:
  - ecosystem-management tooling reuse (repo classification batching, Forgejo remote-URL migration) —
    not for canon/governance content, which is out of scope
  not_recommended_for:
  - any attempt to reuse or reference canon/memory contents (confidential, NDA-protected, explicitly excluded
    from this entry)
  known_limitations:
  - registry entry intentionally narrow; confidentiality boundary must be respected in any future promotion
promotion_history: []
---

# Custodian Runtime And Repo-Classification Tooling

## Overview

The Custodian's repository root is confidential and proprietary. This capability entry deliberately scopes only to its generic, non-confidential tooling surface: the `runtime/` agent framework (context, actions, tool adapters, policies) and the `tools/` repo-classification batch/review/validate/push scripts — including `patch-forgejo-remote-urls.sh`, directly relevant to the ecosystem's Gitea-to-Forgejo transition. Canon and memory content are explicitly excluded from this entry.

## Assessment notes

### Discovery

This repo's README and canon content are marked confidential/proprietary; this entry deliberately scopes to the generic, non-proprietary tooling surface only (runtime/ agent framework, tools/ classification scripts), which is separately reusable and does not disclose canon/memory contents.

### Availability

`runtime/agent.py` plus context/actions/tool_adapters/policies modules form a real Python package (own pyproject, tests, README) under `runtime/`; `tools/` holds standalone scripts (batch_author_repo_classifications.py, human_review_classifications.py, validate_repo_classification.py, push_repo_classifications.sh, patch-forgejo-remote-urls.sh) used for ecosystem-wide repo classification and the Gitea-to-Forgejo remote URL migration.

### Completeness

First-pass honest assessment from the REUSE-WP-0017 coverage campaign
(reuse-surface). No external consumer feedback exists yet; levels reflect
scope-vs-intent documentation quality, not internal code quality.

### Reliability

No production consumer telemetry exists yet; reliability level is
intentionally conservative pending REUSE-WP-0019 reuse-telemetry evidence.

## Promotion checklist

- [x] ID follows `capability.<domain>.<name>` pattern
- [x] Maturity enums match `specs/CapabilityMaturityStandard.md`
- [x] `external_evidence` is populated separately from `maturity`
- [ ] Relations reference valid capability IDs (none yet)
- [x] Index entry added in `registry/indexes/capabilities.yaml`