docs: finalize ADR statuses to Accepted; add ADR README and KSA scaffold

All seven ADRs were in draft/proposed/decided states — promote all to
Accepted now that the decisions are implemented and stable.

Add .claude/cos/adrs/README.md documenting the ADR lifecycle process.

Add KSA Workspace Standards section to CLAUDE.md (sourced from scaffold
at claude-scaffold/CLAUDE.md) so future sessions inherit workspace-wide
rules without needing a separate file load.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Kevin Adams 2026-05-19 11:07:49 -04:00
parent 9876e85d74
commit 4833cfbcca
9 changed files with 159 additions and 7 deletions

View File

@ -1,7 +1,7 @@
# ADR-001: Consolidate Build Pipeline into Main Repo # ADR-001: Consolidate Build Pipeline into Main Repo
**Date**: 2026-05-15 **Date**: 2026-05-15
**Status**: Proposed **Status**: Accepted
**Deciders**: Kevin Adams **Deciders**: Kevin Adams
## Context ## Context

View File

@ -1,7 +1,7 @@
# ADR-002: UI Integration Strategy # ADR-002: UI Integration Strategy
**Date**: 2026-05-15 **Date**: 2026-05-15
**Status**: Under Discussion **Status**: Accepted
**Deciders**: Kevin Adams **Deciders**: Kevin Adams
## Context ## Context

View File

@ -1,7 +1,7 @@
# ADR-003: APT Repository Hosting # ADR-003: APT Repository Hosting
**Date**: 2026-05-15 **Date**: 2026-05-15
**Status**: Under Discussion **Status**: Accepted
**Deciders**: Kevin Adams **Deciders**: Kevin Adams
## Context ## Context

View File

@ -1,7 +1,7 @@
# ADR-004: Transactional Cleanup on API Operation Failure # ADR-004: Transactional Cleanup on API Operation Failure
**Date**: 2026-05-15 **Date**: 2026-05-15
**Status**: Decided **Status**: Accepted
**Deciders**: Kevin Adams **Deciders**: Kevin Adams
## Context ## Context

View File

@ -1,7 +1,7 @@
# ADR-005: Bearer Token Authentication as Primary Auth Method # ADR-005: Bearer Token Authentication as Primary Auth Method
**Date**: 2026-05-15 **Date**: 2026-05-15
**Status**: Decided **Status**: Accepted
**Deciders**: Kevin Adams **Deciders**: Kevin Adams
## Context ## Context

View File

@ -1,7 +1,7 @@
# ADR-006: Package Versioning Strategy # ADR-006: Package Versioning Strategy
**Date**: 2026-05-15 **Date**: 2026-05-15
**Status**: Decided **Status**: Accepted
**Deciders**: Kevin Adams **Deciders**: Kevin Adams
## Context ## Context

View File

@ -1,7 +1,7 @@
# ADR-007: FreeNAS vs TrueNAS Module Naming # ADR-007: FreeNAS vs TrueNAS Module Naming
**Date**: 2026-05-15 **Date**: 2026-05-15
**Status**: Decided **Status**: Accepted
**Deciders**: Kevin Adams **Deciders**: Kevin Adams
## Context ## Context

View File

@ -0,0 +1,32 @@
# Architecture Decision Records
ADRs document significant decisions made about this project. They follow an RFC-style lifecycle.
## Lifecycle
| Status | Meaning |
|--------|---------|
| **Draft** | Being written or discussed. Not yet binding. May change freely. |
| **Accepted** | Agreed upon and in effect. Guides implementation. |
| **Superseded** | Replaced by a newer ADR. The old record is preserved as history. The header notes which ADR supersedes it. |
| **Rejected** | Proposed but explicitly decided against. Preserved so we remember why. |
| **Deprecated** | Was accepted but no longer applicable (e.g., a feature was dropped). Not replaced. |
## Rules
- **Never edit the Decision section of an Accepted ADR.** If the decision changes, write a new ADR and mark the old one `Superseded by ADR-XXX`.
- **Draft ADRs can be edited freely** — they are working documents.
- **Numbering is sequential and permanent.** A superseded ADR keeps its number; the new one gets the next number.
- **Context and Consequences sections** may be lightly edited for clarity even after acceptance (fix typos, add links), but the decision itself is immutable.
## Index
| ADR | Title | Status |
|-----|-------|--------|
| [ADR-001](ADR-001-consolidate-build-into-main-repo.md) | Consolidate Build Pipeline into Main Repo | Accepted |
| [ADR-002](ADR-002-ui-strategy.md) | UI Integration Strategy | Accepted |
| [ADR-003](ADR-003-apt-repo-hosting.md) | APT Repository Hosting | Accepted |
| [ADR-004](ADR-004-cleanup-on-failure.md) | Transactional Cleanup on API Operation Failure | Accepted |
| [ADR-005](ADR-005-bearer-token-authentication.md) | Bearer Token Authentication as Primary Auth Method | Accepted |
| [ADR-006](ADR-006-versioning-strategy.md) | Package Versioning Strategy | Accepted |
| [ADR-007](ADR-007-freenas-vs-truenas-naming.md) | FreeNAS vs TrueNAS Module Naming | Accepted |

120
CLAUDE.md
View File

@ -81,3 +81,123 @@ The plugin auto-detects v1.0 vs v2.0 API based on the TrueNAS version string and
See `.claude/cos/adrs/` for Architecture Decision Records. See `.claude/cos/adrs/` for Architecture Decision Records.
See `.claude/cos/plans/` for implementation plans. See `.claude/cos/plans/` for implementation plans.
See `.claude/cos/runbooks/` for operational runbooks. See `.claude/cos/runbooks/` for operational runbooks.
---
# KSA Workspace Standards
> Baseline rules for all KSA Technologies Claude workspaces.
> Source of truth: `/mnt/c/Users/Kevin/OneDrive - ksatechnologies.com/workspace/claude-scaffold/CLAUDE.md`
## Environment
- **Shell:** WSL (Linux). Always use bash/Linux syntax — `export VAR=value`, forward slashes, `source .venv/activate`. Never PowerShell, never cmd.exe.
- **OS:** WSL2 on Windows. Paths under `/mnt/c/Users/Kevin/OneDrive - ksatechnologies.com/workspace/`.
## Agent / Minion Rules
- **Minions are research-only.** They search, read, and web-fetch. They return findings to main Claude. They never write files, commit, push, or delete.
- **Minions do not talk to each other.** All findings route through main Claude.
- **Main Claude does all writing, committing, and pushing.**
## File Rules
- **New files:** write freely.
- **Changes to existing files:** use targeted edits only — add or modify, never blank out content.
- **Total replacement of an existing file:** write the replacement as a NEW file first, explain to the user why the old file needs replacing, wait for explicit approval. Never silently overwrite.
- **Briefs and reports:** always new dated files. Never overwrite a previous brief. The historical record grows forward, never shrinks.
## Permission Levels
| Actor | Permissions |
|---|---|
| **User** | Full control — can do anything including delete |
| **Main Claude** | Pre-approved: read, write, edit, commit, push, gh CLI. Must ask before any delete unless user has explicitly pre-authorized it. |
| **Minions** | Research only — read, search, web-fetch. No write, commit, push, or delete. |
## Workflow Order
Always follow this sequence. Never skip ahead:
1. **Open a GitHub issue** — captures the problem/feature before any code
2. **Update design/architecture docs** — ADR if warranted, roadmap, architecture
3. **Implement code** — with tests
4. **Update docs** — test-scenarios, user examples, troubleshooting callouts
5. **Close the issue** — with version, commit, and doc pointers
## Documentation Discipline
Every feature, breaking change, or config change ships with:
- Code + tests
- Test-scenario examples
- Troubleshooting callouts where relevant
- Architecture/design doc update if the shape changed
- ADR if a significant technology decision was made
- Roadmap `Resolved` update
Never ship a feature without its docs.
## Issue Lifecycle
- Close issues when work ships — include version, commit SHA, and doc pointers
- User-filed bugs: add `fixed — awaiting-feedback` label if user confirmation needed
- Dependabot patch/minor: merge when CI is green (ad-hoc, no milestone)
- Dependabot major: open an issue, add to a milestone, test intentionally
- Dependabot CVE: link PR to the security issue, include in relevant milestone
## SAFe Hierarchy (GitHub Issues)
| Level | GitHub object | When to use |
|---|---|---|
| **Milestone** | GitHub Milestone | PI / sprint container, ties to a semver release |
| **Feature** | Issue (`enhancement` label + milestone) | Single deployable unit of value — this is the default issue type |
| **Epic** | Issue (`epic` label, no milestone, task list of linked features) | Only when work spans multiple milestones |
| **Story** | Sub-task within a feature | Only when multiple contributors need independently assignable units |
Default: **milestone → feature**. Add epic and story layers only when the team size and cross-milestone scope justify the overhead.
## Technology Governance
Follow the original developer when a project forks due to governance or ownership breakdown. Validated examples:
- pfSense → **OPNsense** (m0n0wall founder Kasper endorsed it)
- CentOS → **Rocky Linux** (co-founder Kurtzer created it within days)
- nginx → watch **freenginx** (core developer Dounin forked over F5 governance)
- OpenOffice → **LibreOffice** (original developers moved there)
**Avoid:** Oracle (hostile OSS track record), F5/nginx (governance broken), Bitnami (stale images since 2024 paid tier split).
**Preferred stack:**
- Server OS: Rocky Linux, Debian
- Container base (community): `python:slim-bookworm`
- Container base (production): `python:alpine`
- Container base (enterprise/FIPS): `cgr.dev/chainguard/python`
- CNI: Cilium
- Ingress: Kubernetes Gateway API (Cilium native implementation)
- Secrets: HashiCorp Vault sidecar / CSI driver
- Postgres (single node): `postgres:16-alpine` custom subchart
- Postgres (HA): CloudNativePG operator
## ADR Process
ADRs follow an RFC lifecycle: **Draft → Proposed → Accepted / Rejected / Withdrawn**. When a decision changes, a new ADR is created and the old one is marked **Superseded** — the historical record is never deleted.
See `ADR-000-process.md` in any project that uses ADRs for the full format and lifecycle definition.
## Git Commit Style
- Conventional commits: `feat:`, `fix:`, `docs:`, `chore:`, `ci:`, `refactor:`
- Co-author line on every commit: `Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>`
- Commit message: why, not what. The diff shows what; the message explains why.
- Never amend published commits. Create new commits instead.
- Never skip hooks (`--no-verify`) unless explicitly instructed.
## Memory System
Memory lives at: `.claude/projects/<workspace-path>/memory/MEMORY.md`
On starting a new session, read `MEMORY.md` first. It indexes all project-specific memories. Update memory when:
- User corrects an approach or confirms a non-obvious one
- A significant project decision is made
- User preferences or constraints are stated
Never save ephemeral task details to memory — use the todo list for that.