docs: sync agent guide with AGENTS

This commit is contained in:
Tam Nhu Tran
2026-05-19 18:49:33 -04:00
parent 8c8f93a67e
commit b123d41590
+86 -525
View File
@@ -1,563 +1,124 @@
# Agent Guidelines
# CCS CLI Agent Guide
AI-facing guidance for agent tooling when working with this repository.
Canonical agent instructions for `/Users/kaitran/CloudPersonal/ccs/cli`.
`AGENTS.md` must stay a symlink to this file.
## Critical Constraints (NEVER VIOLATE)
## Scope
### Test Isolation (MANDATORY)
CCS is a TypeScript/Bun CLI and dashboard for managing Claude Code, Codex,
Factory Droid, CLIProxy, and compatible provider profiles.
**NEVER touch the user's real `~/.ccs/` or `~/.claude/` directories during tests.**
## Non-Negotiables
- All code accessing CCS paths MUST use `getCcsDir()` from `src/utils/config-manager.ts`
- This function respects `CCS_HOME` env var for test isolation
- **WRONG:** `path.join(os.homedir(), '.ccs', ...)`
- **CORRECT:** `path.join(getCcsDir(), ...)`
- Default branch is `dev`. Feature/fix branches start from `dev`; production
hotfixes start from `main` only when explicitly needed.
- Never touch the user's real `~/.ccs/` or `~/.claude/` in tests. Use
`getCcsDir()` from `src/utils/config-manager.ts`; it respects `CCS_HOME`.
- Do not commit directly to `dev` or `main`.
- Do not manually bump versions or create release tags. Semantic-release owns
versions, changelog, tags, npm publish, and GitHub releases.
- CLI terminal output must be ASCII only: `[OK]`, `[!]`, `[X]`, `[i]`.
- Respect `NO_COLOR` and TTY-aware output.
Tests set `process.env.CCS_HOME` to a temp directory. Code using `os.homedir()` directly will modify the user's real files.
## Architecture
## CI-First Protocol (MANDATORY)
- `src/` - TypeScript CLI/server source.
- `lib/ccs`, `lib/ccs.ps1` - bootstrap wrappers; no help text here.
- `ui/src/` - React dashboard.
- `dist/` and `dist/ui/` - build outputs.
- `docs/` - local development and architecture docs.
- Docker support lives under `docker/` and related commands.
**A task is NOT complete until CI is green. After every `git push`, the AI agent MUST block on CI until it passes.**
Profile resolution priority:
### Required Sequence
1. `git push`
2. **Immediately** run `gh pr checks --watch` (or `gh run watch`) and block until all checks complete.
3. If **green** → task may proceed to next step / be declared done.
4. If **red**:
- Pull failing logs: `gh run view --log-failed` (or `gh pr checks <n>` to identify the failing job, then `gh run view <run-id> --log-failed`).
- Fix the root cause locally. Do NOT retry blindly.
- Commit and push again. Re-watch CI.
5. Applies to initial `gh pr create` AND every subsequent push on an open PR.
1. Built-in CLIProxy providers: Gemini, Codex, Antigravity.
2. User-defined `config.cliproxy` providers.
3. Settings-based `config.profiles`.
4. Account-based `profiles.json` with isolated `CLAUDE_CONFIG_DIR`.
### Fallback (when `--watch` is unavailable or flaky)
Poll with short sleep until no check is `pending` / `in_progress`:
```bash
until [ "$(gh pr checks <n> --json state -q '[.[] | select(.state == "IN_PROGRESS" or .state == "PENDING" or .state == "QUEUED")] | length')" = "0" ]; do
sleep 10
done
gh pr checks <n>
```
All env values written into settings must be strings.
### Absolute rule
AI MUST NOT declare a task done, close a session, or move to the next task while CI is red or still running. Leaving a PR red and moving on is the primary failure mode this protocol prevents.
## User-Facing Change Checklist
### Dev Release vs Push CI
- Update the matching `--help` handler when CLI behavior changes.
- Keep README concise; do not remove `## Community Projects` or
`## Star History` unless explicitly asked.
- Use neutral broad examples such as `ccs`, `ccs codex`, `ccs glm`, or
`ccs <provider>` unless the page is provider-specific.
- If CLI commands, config, providers, install steps, or user workflows change,
update the public CCS docs in `/Users/kaitran/CloudPersonal/ccs/docs`.
- `CI` is the pull-request quality gate for contributor branches.
- `Push CI` is the post-merge quality signal for `dev`.
- `Dev Release` publishes the `@dev` package after `dev` changes land.
- A red `Dev Release` does **not** automatically mean contributor code failed. Check `Push CI` first.
- Verified on `2026-04-22` via `gh api repos/kaitranntt/ccs/branches/dev/protection`: `dev` currently requires `typecheck`, `lint`, `format`, `build`, and `test`, has no branch restrictions, and has no required PR-review gate.
- `dev-release.yml` currently pushes with `PAT_TOKEN` because `dev` is protected by those required status checks. Do not switch it back to `github.token` unless branch protection changes with it.
Help locations:
### Self-Hosted Runner Policy
- `ccs --help`: `src/commands/help-command.ts`
- `ccs api --help`: `src/commands/api-command.ts`
- `ccs cleanup --help`: `src/commands/cleanup-command.ts`
- `ccs cliproxy --help`: `src/commands/cliproxy-command.ts`
- `ccs config --help`: `src/commands/config-command.ts`
- `ccs copilot --help`: `src/commands/copilot-command.ts`
- `ccs cursor --help`: `src/commands/cursor-command.ts`
- `ccs doctor --help`: `src/commands/doctor-command.ts`
- `ccs docker --help`: `src/commands/docker/help-subcommand.ts`
- `ccs env --help`: `src/commands/env-command.ts`
- `ccs migrate --help`: `src/commands/migrate-command.ts`
- `ccs persist --help`: `src/commands/persist-command.ts`
- `ccs setup --help`: `src/commands/setup-command.ts`
- Keep active CCS workflows on local self-hosted runners by default. Do **not** move jobs to `ubuntu-latest`, `macos-latest`, or other GitHub-hosted runners as the first security fix.
- For this public repo, standard GitHub-hosted runners may not consume billable minutes, but the project policy is still local-runner-first to keep compute predictable and avoid future quota surprises across repos.
- Make self-hosted use secure by gating untrusted PR triggers, avoiding checkout of untrusted code in privileged `pull_request_target` jobs, setting `persist-credentials: false`, and scoping PAT-backed credentials to the exact release/sync push step.
- If a workflow truly needs GitHub-hosted runners, document the exception in this file and add a regression test for the exception.
## Validation
## Core Function
Multi-provider profile and runtime manager for Claude Code, Factory Droid,
Codex CLI, and other compatible targets. See README.md for user documentation.
## README Preservation
When editing `README.md`, keep the file concise and funnel detailed usage into
the docs site, but **do not remove the `## Community Projects` section** or the
`## Star History` section unless the user explicitly asks for those sections to
be deleted. Treat both as protected README content.
When a contributor adds a useful community integration section to `README.md`,
prefer preserving the attribution in `## Community Projects` and moving the
setup substance into a docs page, rather than deleting the contribution.
Outside provider-specific Gemini and Antigravity docs, avoid using `ccs gemini`
or `ccs agy` as the primary hero example, default starter route, or generic
workflow example. Prefer `ccs`, `ccs codex`, `ccs kiro`, `ccs glm`, Droid
examples, or neutral `ccs <provider>` placeholders when the page is about a
broader topic.
## Design Principles (ENFORCE STRICTLY)
### Technical Excellence
- **YAGNI**: No features "just in case"
- **KISS**: Simple bash/PowerShell/Node.js only
- **DRY**: One source of truth (config.yaml)
### User Experience (EQUALLY IMPORTANT)
- **CLI-Complete**: All features MUST have CLI interface
- **Dashboard-Parity**: Configuration features MUST also have Dashboard interface
- **Execution is CLI**: Running profiles happens via terminal, not dashboard buttons
- **UX > Brevity**: Error messages and help text prioritize user success over terseness
- **Progressive Disclosure**: Simple by default, power features accessible but not overwhelming
### When Principles Conflict
- **UX > YAGNI** for user-facing features (if users need it, it's not "just in case")
- **KISS applies to BOTH** code AND user experience (simple journey, not just simple code)
- **DRY applies to BOTH** code AND interface patterns (consistent behavior across CLI/Dashboard)
## Common Mistakes (AVOID)
| Mistake | Consequence | Correct Action |
|---------|-------------|----------------|
| Running `validate` without `format` first | format:check fails | Run `bun run format` BEFORE validate |
| Treating `Dev Release` as the contributor quality signal | Publish failures on `dev` look like broken code | Check PR `CI` on the branch and `Push CI` on `dev` first |
| Using `chore:` for dev→main PR | No npm release triggered | Use `feat:` or `fix:` prefix |
| Committing directly to `main` or `dev` | Bypasses CI/review | Always use PRs |
| Manual version bump or git tag | Conflicts with semantic-release | Let CI handle versioning |
| Forgetting `--help` update | CLI docs out of sync | Update `src/commands/help-command.ts` |
| Forgetting docs update | User docs out of sync | Update `docs/` and CCS docs submodule |
## GitHub Issue Operations (CCS-Specific)
These rules apply when the task is issue triage, backlog cleanup, labels, comments, Projects, or milestones for this repo.
### Scope Boundary
- Treat issue triage as a **GitHub-only workflow** unless the user explicitly asks for implementation.
- Do **NOT** create a worktree, branch, PR, or run `/fix`, `/cook`, or `kai:maintainer` just to tag issues, post follow-up comments, close duplicates, or clean up backlog state.
- Escalate into code workflow only when:
- the user explicitly asks to fix/implement an issue, or
- triage proves the same task now requires code changes.
### Read Before Mutating
- Always inspect live issue state first with `gh issue view <n> --json ...` or `gh api`.
- Never rely on stale memory, screenshots, or issue titles alone.
- Before closing as resolved, cross-check repo evidence in at least one of:
- `README.md`
- `docs/`
- `CHANGELOG.md`
- relevant source/help handlers
- If the `gh` query would touch Projects fields, verify token scope first. Missing `read:project` is a real blocker, not something to hand-wave around.
### Labeling Standard
- Every **open** issue should end triage with:
- one primary type label: `bug`, `enhancement`, `question`, `documentation`, `duplicate`, `invalid`, or `wontfix`
- one area label:
- `area:cli-runtime`
- `area:dashboard-ui`
- `area:config-auth`
- `area:provider-integration`
- `area:install-packaging`
- `area:documentation`
- `area:contributor-workflow`
- Add routing labels only when they materially change handling:
- `upstream-blocked`
- `needs-repro`
- `needs-split`
- `docs-gap`
- Use release-state labels for shipped work:
- `pending-release`
- `released-dev`
- `released`
- Do **NOT** create or use status labels like `todo`, `doing`, `blocked`, `done`.
- Do **NOT** create provider-name labels unless there is a proven long-term need. Provider names belong in titles/issues, not label spam.
### Commenting Rules
- Keep issue comments short, technical, and neutral.
- State the decision plainly: close, keep open, retag, needs repro, duplicate, blocked upstream.
- Include exact evidence when relevant: version, doc path, changelog release, canonical issue, upstream link.
- Do **NOT** reference internal plans, local report files, agent prompts, or private reasoning.
- Post **one** maintainer follow-up comment per triage pass. If accidental duplicates are created, delete them with `gh api repos/<owner>/<repo>/issues/comments/<id> -X DELETE`.
### Closure Rules
- Close immediately when:
- the issue is an obvious duplicate and you can point to the canonical issue
- the feature/fix is clearly shipped and documented
- a previously `pending-release` issue is now clearly past release and no longer needs tracking
- Keep open and retag when:
- upstream dependency still blocks CCS adoption -> `upstream-blocked`
- latest-release behavior is unclear -> `needs-repro`
- issue contains multiple independent asks -> `needs-split`
- feature likely exists but discoverability/docs are weak -> `docs-gap`
- Do **NOT** close just because an issue is old, vague, or inconvenient. Close only with evidence.
### Projects And Milestones
- Preferred project model for this repo: one project, `CCS Backlog`.
- Use Projects for workflow state and priority. Use labels for meaning and routing.
- Milestones are for real ship windows only, not generic categorization buckets.
- If `gh` token lacks `read:project`, say so explicitly and stop short of pretending Projects data is available.
- Active project:
- owner: `kaitranntt`
- number: `3`
- URL: `https://github.com/users/kaitranntt/projects/3`
- Active project fields:
- `Status` -> use for work state (`Todo`, `In Progress`, `Done`)
- `Priority` -> `P1` for bugs, `P2` default backlog, `P3` for broad `needs-split` buckets unless explicitly reprioritized
- `Follow-up` -> `Ready`, `Needs repro`, `Blocked upstream`, `Needs split`, `Docs follow-up`
- `Next review` -> date only for issues that need a follow-up checkpoint
- When triaging an open issue, make sure it exists in `CCS Backlog` and the project fields match the routing labels.
- Do **NOT** create a second backlog project unless the user explicitly wants a project split and gives a reason.
- Current automation path:
- workflow file: `.github/workflows/sync-ccs-backlog-project.yml`
- sync script: `scripts/github/ccs-backlog-sync.mjs`
- required Actions secret: `CCS_PROJECT_AUTOMATION_TOKEN`
- Automation mapping must stay aligned with labels:
- `upstream-blocked` -> `Follow-up=Blocked upstream`
- `needs-repro` -> `Follow-up=Needs repro`
- `needs-split` -> `Follow-up=Needs split`
- `docs-gap` -> `Follow-up=Docs follow-up`
- otherwise -> `Follow-up=Ready`
### New Or Updated Issue Creation
- When creating issues for this repo:
- assign `@kaitranntt`
- use conventional issue titles: `bug: ...`, `feat: ...`, `docs: ...`
- keep bodies factual and technical
- avoid personal info and internal-only context
## Quality Gates (MANDATORY)
Quality gates MUST pass before pushing. **Both projects have identical workflow.**
### Pre-Commit Sequence (FOLLOW THIS ORDER)
Format before validating:
```bash
# Main project (from repo root)
bun run format # Step 1: Fix formatting
bun run lint:fix # Step 2: Fix lint issues
bun run validate # Step 3: Fast gate (typecheck + lint + format + test:fast)
bun run validate:ci-parity # Step 4: PR-CI parity gate (branch check + build + full tests + e2e)
# UI project (if UI changed)
cd ui
bun run format # Step 1: Fix formatting
bun run lint:fix # Step 2: Fix lint issues
bun run validate # Step 3: Final check (must pass)
cd /Users/kaitran/CloudPersonal/ccs/cli && bun run format
cd /Users/kaitran/CloudPersonal/ccs/cli && bun run lint:fix
cd /Users/kaitran/CloudPersonal/ccs/cli && bun run validate
```
**WHY THIS ORDER:**
- `validate` runs `format:check` which only VERIFIES—won't fix
- If format:check fails, you skipped step 1
- `validate` now uses read-only `lint`, so autofix still belongs in step 2
- PR CI and `validate:ci-parity` both run non-mutating checks only
Before requesting review or merge, run:
### What Each Gate Runs
| Project | Command | Runs |
|---------|---------|------|
| Main | `bun run validate` | typecheck + lint + format:check + test:fast |
| Main | `bun run validate:ci-parity` | base branch check + typecheck + lint + format:check + build:all + test:all + test:e2e |
| UI | `bun run validate` | typecheck + lint:fix + format:check |
### ESLint Rules (ALL errors)
| Rule | Level | Notes |
|------|-------|-------|
| `@typescript-eslint/no-unused-vars` | error | Ignore `_` prefix |
| `@typescript-eslint/no-explicit-any` | error | Use proper types or `unknown` |
| `@typescript-eslint/no-non-null-assertion` | error | No `!` assertions |
| `prefer-const`, `no-var`, `eqeqeq` | error | Code quality |
| `react-hooks/*` (UI only) | recommended | Hooks rules |
| `react-refresh/*` (UI only) | vite | Fast refresh |
### TypeScript Options (strict mode)
| Option | Value | Notes |
|--------|-------|-------|
| `strict` | true | All strict flags enabled |
| `noUnusedLocals` | true | No unused variables |
| `noUnusedParameters` | true | No unused params |
| `noImplicitReturns` | true | All paths must return |
| `noFallthroughCasesInSwitch` | true | Explicit case handling |
### Automatic Enforcement
- `prepack` runs `build:all`
- PR `CI` runs `typecheck`, `lint`, `format`, `build`, `test:all`, and `test:e2e`
- `Push CI` runs the same quality suite on `dev` after merge, separate from release publishing
- `Dev Release` still runs build + fast validation + slow tests + e2e before publishing and still requires `PAT_TOKEN` to push back to protected `dev`
- husky `pre-commit` runs quick lint/type/format checks
- husky `pre-push` runs the full `bun run validate:ci-parity` gate on `main`/`dev`/hotfix branches
- husky `pre-push` runs a faster feature-branch gate (`typecheck` + `lint` + `format:check` + `test:fast`) plus targeted checks based on changed files
### Maintainability Gate Status
- The historical maintainability baseline gate is retired from the active CCS workflow.
- `validate`, `validate:ci-parity`, PR `CI`, `Push CI`, and release workflows do **not** invoke `maintainability:check`.
- Older roadmap references to `maintainability:baseline` / `maintainability:check` are historical context, not current repo commands.
## Critical Constraints (NEVER VIOLATE)
1. **NO EMOJIS in CLI output** - Terminal output uses ASCII only: [OK], [!], [X], [i]
- **Scope:** CCS CLI terminal output (`src/` code that prints to stdout/stderr)
- **Does NOT apply to:** PR descriptions, commit messages, documentation, comments, AI conversations
2. **TTY-aware colors** - Respect NO_COLOR env var
3. **Non-invasive** - NEVER modify external tool settings (`~/.claude/settings.json`) without explicit user request and confirmation (exception: `ccs persist` command)
4. **Cross-platform parity** - bash/PowerShell/Node.js must behave identically
5. **CLI documentation** - ALL CLI changes MUST update respective `--help` handler (see table below)
6. **Idempotent** - All install operations safe to run multiple times
7. **Dashboard parity** - Configuration features MUST work in both CLI and Dashboard
### Help Location Reference
| Command | Help Handler Location |
|---------|----------------------|
| `ccs --help` | `src/commands/help-command.ts` |
| `ccs api --help` | `src/commands/api-command.ts``showHelp()` |
| `ccs cleanup --help` | `src/commands/cleanup-command.ts``printHelp()` |
| `ccs cliproxy --help` | `src/commands/cliproxy-command.ts``showHelp()` |
| `ccs config --help` | `src/commands/config-command.ts``showHelp()` |
| `ccs copilot --help` | `src/commands/copilot-command.ts``handleHelp()` |
| `ccs cursor --help` | `src/commands/cursor-command.ts``handleHelp()` |
| `ccs doctor --help` | `src/commands/doctor-command.ts``showHelp()` |
| `ccs docker --help` | `src/commands/docker/help-subcommand.ts``showHelp()` |
| `ccs migrate --help` | `src/commands/migrate-command.ts``printMigrateHelp()` |
| `ccs env --help` | `src/commands/env-command.ts``showHelp()` |
| `ccs persist --help` | `src/commands/persist-command.ts``showHelp()` |
| `ccs setup --help` | `src/commands/setup-command.ts``showHelp()` |
**Note:** `lib/ccs` and `lib/ccs.ps1` are bootstrap wrappers only—they delegate to Node.js and contain no help text.
## Documentation Requirements (MANDATORY)
**Documentation is a first-class citizen. ALL user-facing changes require docs updates.**
### Local Documentation (`docs/`)
Update local `docs/` folder for:
- Architecture changes
- Internal API documentation
- Development guides
### CCS Docs Submodule (Owner Only)
**For @kaitranntt (repository owner):** When adding/changing CLI commands or config options, you MUST also update the CCS docs submodule at `~/CloudPersonal/ccs/docs/`:
| Change Type | Files to Update |
|-------------|-----------------|
| New CLI command/flag | `reference/cli-commands.mdx` |
| New config option | `reference/config-schema.mdx` |
| Provider feature | `providers/<provider>.mdx` |
| New feature | `features/<feature>.mdx` |
**Workflow for docs submodule:**
```bash
cd ~/CloudPersonal/ccs/docs/
git checkout main && git pull
# Make changes
git add -A && git commit -m "docs: <description>"
git push origin main
cd /Users/kaitran/CloudPersonal/ccs/cli && bun run validate:ci-parity
```
**For external contributors:** Document changes in PR description. Owner will sync to CCS docs.
If UI changed:
### Pre-Commit Docs Checklist
- [ ] Respective `--help` updated (see Help Location Reference table)
- [ ] Local `docs/` updated if architecture changed
- [ ] CCS docs submodule updated (owner) or PR description includes docs (contributor)
## Feature Interface Requirements
| Feature Type | CLI | Dashboard | Example |
|--------------|-----|-----------|---------|
| Profile creation | ✓ | ✓ | `ccs auth create`, Dashboard "Add Account" |
| Profile switching | ✓ | ✓ | `ccs <profile>` (execution is CLI-only) |
| API key config | ✓ | ✓ | `ccs api create`, Dashboard API Profiles |
| Health check | ✓ | ✓ | `ccs doctor`, Dashboard Live Monitor |
| OAuth auth flow | ✓ | ✓ | Browser opens from CLI or Dashboard |
| Analytics/monitoring | ✗ | ✓ | Dashboard Analytics (visual by nature) |
| WebSearch config | ✓ | ✓ | CLI flags, Dashboard Settings |
| Remote proxy config | ✓ | ✓ | CLI flags, Dashboard Settings |
## File Structure
```
src/ → TypeScript source (main project)
dist/ → Compiled JavaScript (npm package)
lib/ → Native shell scripts (bash, PowerShell)
ui/src/ → React components, hooks, pages
ui/src/components/ui/ → shadcn/ui components
dist/ui/ → Built UI bundle (served by Express)
```
## Key Technical Details
### Profile Mechanisms (Priority Order)
1. **CLIProxy hardcoded**: gemini, codex, agy → OAuth-based, zero config
2. **CLIProxy variants**: `config.cliproxy` section → user-defined providers
3. **Settings-based**: `config.profiles` section → GLM, legacy GLMT compatibility, Kimi
4. **Account-based**: `profiles.json` → isolated instances via `CLAUDE_CONFIG_DIR`
### Settings Format (CRITICAL)
All env values MUST be strings (not booleans/objects) to prevent PowerShell crashes.
```json
{
"env": {
"ANTHROPIC_BASE_URL": "https://api.example.com/anthropic",
"ANTHROPIC_AUTH_TOKEN": "your-api-key",
"ANTHROPIC_MODEL": "model-name"
}
}
```
### Shared Data Architecture
Symlinked from `~/.ccs/shared/`: commands/, skills/, agents/
Profile-specific: settings.json, sessions/, todolists/, logs/
Windows fallback: Copies if symlinks unavailable
## Code Standards
### Architecture
- `lib/ccs`, `lib/ccs.ps1` - Bootstrap scripts (delegate to Node.js via npx)
- `src/*.ts``dist/*.js` - Main implementation (TypeScript)
### Bash (lib/*.sh)
- bash 3.2+, `set -euo pipefail`, quote all vars `"$VAR"`, `[[ ]]` tests
- NO external dependencies
### PowerShell (lib/*.ps1)
- PowerShell 5.1+, `$ErrorActionPreference = "Stop"`
- Native JSON only, no external dependencies
### TypeScript (src/*.ts)
- Node.js 18+, Bun 1.0+, TypeScript 5.3, strict mode
- `child_process.spawn`, handle SIGINT/SIGTERM
### Terminal Output
- ASCII only: [OK], [!], [X], [i] (NO emojis in CLI output)
- TTY detect before colors, respect NO_COLOR
- Box borders for errors: ╔═╗║╚╝
## Conventional Commits (MANDATORY)
**ALL commits MUST follow conventional commit format. Non-compliant commits are rejected by husky.**
### Format
```
<type>(<scope>): <description>
```
### Types (determines version bump)
| Type | Version Bump | Use For |
|------|--------------|---------|
| `feat:` | MINOR | New features |
| `fix:` | PATCH | Bug fixes |
| `perf:` | PATCH | Performance |
| `feat!:` | MAJOR | Breaking changes |
| `docs:`, `style:`, `refactor:`, `test:`, `chore:`, `ci:`, `build:` | None | Non-release |
### Examples
```bash
# Good
git commit -m "feat(cliproxy): add OAuth token refresh"
git commit -m "fix(doctor): handle missing config gracefully"
# Bad - REJECTED
git commit -m "added new feature"
git commit -m "Fixed bug"
cd /Users/kaitran/CloudPersonal/ccs/cli/ui && bun run format && bun run validate
```
## Branching Strategy
After every push to a PR, watch CI until it finishes. If checks fail, inspect
logs, fix root cause, push again, and re-watch.
### Hierarchy
```
main (production) ← dev (integration) ← feat/* | fix/* | docs/*
└── hotfix/* (critical only, skips dev)
```
## Issue Triage
Issue triage is GitHub-only unless implementation is explicitly requested.
Always inspect live state first:
### Standard Workflow
```bash
git checkout dev && git pull origin dev
git checkout -b feat/my-feature
# ... develop with conventional commits ...
git push -u origin feat/my-feature
gh pr create --base dev --title "feat(scope): description"
# After testing in @dev:
gh pr create --base main --title "feat(release): promote dev to main"
cd /Users/kaitran/CloudPersonal/ccs/cli && gh issue view <number> --json title,body,state,labels,assignees,comments
```
### Hotfix Workflow (Production Emergencies Only)
```bash
git checkout main && git pull origin main
git checkout -b hotfix/critical-bug
# ... fix ...
gh pr create --base main --title "fix: critical issue"
# Then sync: git checkout dev && git merge main && git push
```
For open issues, prefer one type label and one area label. Use routing labels
only when they affect handling: `upstream-blocked`, `needs-repro`,
`needs-split`, `docs-gap`. Do not close issues on age, intuition, or vague
titles; close only with evidence from README, docs, changelog, source, or a
canonical duplicate.
### Rules
1. **NEVER** commit directly to `main` or `dev`
2. Feature branches from `dev`, hotfixes from `main`
3. dev→main PRs MUST use `feat:` or `fix:` (not `chore:`)
4. Delete branches after merge
## Release Signals
## Automated Releases (DO NOT MANUALLY TAG)
- PR `CI` is the contributor quality gate.
- `Push CI` is the post-merge signal for `dev`.
- `Dev Release` publishes the `@dev` npm package.
- A red `Dev Release` is not automatically contributor failure; check PR `CI`
and `Push CI` first.
**Releases are FULLY AUTOMATED via semantic-release. NEVER manually bump versions or create tags.**
Use `feat:` or `fix:` for dev-to-main promotion PRs so release automation runs.
| Branch | npm Tag | When |
|--------|---------|------|
| `main` | `@latest` | Merge PR to main |
| `dev` | `@dev` | Push to dev branch |
## Design Standards
**CI handles:** version bump, CHANGELOG.md, git tag, npm publish, GitHub release.
## Development
### Testing (REQUIRED before PR)
```bash
bun run test # All tests
bun run test:npm # npm package tests
bun run test:native # Native install tests
bun run test:unit # Unit tests
```
### Local Development
```bash
bun run dev # Build + start config server (http://localhost:3000)
bun run dev:symlink # Symlink global 'ccs' → dev dist/ccs.js (fast iteration)
bun run dev:unlink # Restore original global ccs
./scripts/dev-install.sh # Build, pack, install globally (full install)
rm -rf ~/.ccs # Clean environment
```
**IMPORTANT:** Use `bun run dev` at CCS root for always up-to-date code. Do NOT use `ccs config` during development as it uses the globally installed version.
## Two-Tier Pre-Push Checklist
Optimized for iterative push-then-review workflow. Do NOT run the full gate on every push — CI is the safety net. Run the full gate once before asking for review / merge.
### Tier 1 — Iterative push (feature branch)
Husky `pre-push` auto-runs: `typecheck + lint + format:check + test:fast` plus targeted checks based on changed files. AI does **nothing extra** at push time.
**After push (MANDATORY):** follow the [CI-First Protocol](#ci-first-protocol-mandatory) — watch CI until green. Do not move on while CI is red.
### Tier 2 — Before requesting review / merge
Run ONCE, not per push:
- [ ] `bun run validate:ci-parity` — branch freshness + build + full non-e2e tests + e2e
- [ ] `gh pr checks <n>` — all checks green
- [ ] If UI changed: `cd ui && bun run format && bun run validate`
- [ ] If touching command routing, proxy flows, workflows, or release logic: `bun run test:e2e`
### Code / Docs / Standards (verify before merge)
- [ ] Conventional commit format (`feat:`, `fix:`, etc.)
- [ ] Respective `--help` updated (see Help Location Reference) — if CLI changed
- [ ] Tests added/updated — if behavior changed
- [ ] README.md updated — if user-facing
- [ ] CCS docs updated (owner: `~/CloudPersonal/ccs/docs/`) — if CLI/config changed
- [ ] Local `docs/` updated — if architecture changed
- [ ] CLI output ASCII only (NO emojis in terminal output), NO_COLOR respected
- [ ] YAGNI/KISS/DRY alignment verified
- [ ] No manual version bump or tags
## Error Handling Principles
- Validate early, fail fast with clear messages
- Show available options on mistakes
- Never leave broken state
- YAGNI, KISS, DRY.
- CLI-complete: core configuration features need CLI coverage.
- Dashboard parity: configuration features usually need dashboard coverage too.
- Execution remains CLI-first; dashboard should not replace terminal profile
launch flows.
- Error messages should help users recover, not just report failure.