meto-cli 0.9.3 → 0.11.0

package/templates/.claude/agents/pm-agent.md

@@ -47,7 +47,7 @@ Never write to `/src/`, `tasks-in-progress.md`, `tasks-in-testing.md`, or `tasks
 ## Context Budget
 - Grep before reading — only open files you need
 - Use targeted line ranges for long files
-- Max 3 files open before acting — note key info in memory
+- Max 10 files open before acting — note key info in memory
 - Read each epic once, then work from memory — don't re-read `/ai/backlog/epics.md` every task
 
 ## Task Definition Format
@@ -85,4 +85,4 @@ Sprint [N] is ready. [X] tasks in tasks-todo.md.
 
 When all slices of an epic are in `tasks-done.md`, tell the user:
 
-Epic [N] complete! Consider starting a new session for the next epic — fresh context keeps agents sharp.
+Epic [N] complete! You can continue to the next epic or start a fresh session if context feels heavy.

package/templates/.claude/agents/tester-agent.md

@@ -10,7 +10,8 @@ tools: Read, Bash, Glob, Grep
 1. Read `CLAUDE.md`
 2. Read `.claude/agent-memory/meto-tester/MEMORY.md`
 3. Read `/ai/workflows/definition-of-done.md`
-4. Proceed with validation
+4. Read `/ai/workflows/code-guidelines.md` — verify these during validation
+5. Proceed with validation
 
 ## Session End
 Update `.claude/agent-memory/meto-tester/MEMORY.md` with patterns worth remembering.
@@ -38,7 +39,7 @@ Process items sequentially even when other agents run in parallel.
 ## Context Budget
 - Grep before reading — only open files you need
 - Use targeted line ranges for long files
-- Max 3 files open before acting — note key info in memory
+- Max 10 files open before acting — note key info in memory
 - Only read files listed in "Files changed" on the task block — don't explore the whole codebase
 
 ## Validation Protocol
@@ -61,6 +62,8 @@ ONE item at a time — parallel writes corrupt the board. Always sequential.
 - [ ] Error states handled
 - [ ] No hardcoded secrets
 - [ ] No broken imports
+- [ ] No file exceeds 300 lines (check changed files with `wc -l`)
+- [ ] No function exceeds 50 lines
 
 ## Pass Note
 ```

package/templates/CLAUDE.md

@@ -24,10 +24,10 @@ cd your-project && claude          # start a session
 Esc Esc → /rewind                  # undo if something went wrong
 ```
 
-- **One session per epic/feature** — start fresh when you finish an epic
+- **One session per epic/feature** — start fresh when you finish an epic or context feels heavy
 - **Commit frequently** — don't let work pile up uncommitted
-- **`/compact` early** — don't wait for context to degrade, compress proactively
-- **New session signs:** sluggish responses, repeated file reads, agent forgetting prior work
+- **`/compact` when needed** — context is large (1M tokens), but compress if responses slow down
+- **New session signs:** agent forgetting prior work, repeated file reads, sluggish responses
 
 ## Quick Reference
 
@@ -60,7 +60,7 @@ See `/ai/workflows/definition-of-done.md` for done criteria.
 
 ## Agent Teams
 
-This project supports Agent Teams. The lead agent can spawn teammates using `@meto-pm`, `@meto-developer`, `@meto-tester`.
+This project supports Agent Teams. The lead agent can spawn teammates using `@meto-pm`, `@meto-developer`, `@meto-tester`, `@meto-community`.
 
 **Coordination model:** Agent Teams has its own task system, but this project uses the kanban board (`tasks-backlog` through `tasks-done`) as the single source of truth for task state. Teammates must read and update the board files, not rely on Agent Teams' internal task tracking.
 
@@ -71,6 +71,7 @@ This project supports Agent Teams. The lead agent can spawn teammates using `@me
 | `@meto-pm` | `/ai/` files, `tasks-backlog.md`, `tasks-todo.md` |
 | `@meto-developer` | `/src/`, config files, `tasks-in-progress.md`, `tasks-in-testing.md` |
 | `@meto-tester` | `tasks-in-testing.md`, `tasks-done.md`, `tasks-todo.md` (failed items) |
+| `@meto-community` | Read-only — reads product files, drafts community content in `/ai/community/` |
 
 Each agent writes only its own memory file in `.claude/agent-memory/` -- never another agent's.
 
@@ -80,10 +81,10 @@ Teammates do NOT inherit the lead's conversation history. Each teammate reads CL
 
 ## Context Management
 
-- **Session cadence:** Start a new session every 3-5 slices or when context feels sluggish
+- **Session cadence:** Start a new session every 10-15 slices or when context feels sluggish
 - **Session start:** Read CLAUDE.md, your agent memory file, and the board — then act
 - **Session end:** Update your memory file with decisions, patterns, and what to pick up next
-- **Context budget:** Grep before reading full files; read targeted line ranges; max 3 files open before acting
+- **Context budget:** Grep before reading full files; read targeted line ranges; max 10 files open before acting
 - **Red flag:** If you re-read a file you already read this session, note key info in memory instead
 
 ---
@@ -107,6 +108,9 @@ chore(scope): description [bootstrap]
 <!-- Keep to 3-4 rules max. More rules = more ignored. Focus on what actually causes bugs. -->
 {{CODE_CONVENTIONS}}
 
+## Code Guidelines
+See `/ai/workflows/code-guidelines.md` for file size limits, structure rules, naming, and stack-specific standards. Both `@meto-developer` and `@meto-tester` enforce these.
+
 ---
 
 ## Buildrack — Project Tracking

package/templates/ai/workflows/code-guidelines.md

@@ -0,0 +1,46 @@
+# Code Guidelines — {{PROJECT_NAME}}
+
+These guidelines are enforced by `@meto-developer` during implementation and verified by `@meto-tester` before sign-off.
+
+## File Size Limits
+
+| Scope | Max Lines | Action |
+|---|---|---|
+| Single file | 300 | Split into focused modules |
+| Single function/method | 50 | Extract helper functions |
+| Single React component | 200 | Break into sub-components |
+
+If a file exceeds these limits, refactor before adding more code. Never let a file grow past 500 lines — that is a hard stop.
+
+## Structure Rules
+
+- **One responsibility per file** — if a file does two things, split it
+- **No god files** — a file that everything imports from is a design smell
+- **No circular imports** — if A imports B and B imports A, introduce a shared module or restructure
+- **Flat over nested** — prefer 2 levels of directory nesting max; deeply nested trees are hard to navigate
+- **Co-locate related code** — tests next to source, types next to usage
+
+## Naming
+
+- Files: `kebab-case.ts` (e.g., `user-service.ts`)
+- Classes/Types: `PascalCase` (e.g., `UserService`)
+- Functions/Variables: `camelCase` (e.g., `getUserById`)
+- Constants: `UPPER_SNAKE_CASE` (e.g., `MAX_RETRIES`)
+- Boolean variables: prefix with `is`, `has`, `should` (e.g., `isLoading`)
+
+## Code Hygiene
+
+- No `console.log` in committed code — use a proper logger
+- No commented-out code — delete it, git has history
+- No `any` types (TypeScript) — use `unknown` and narrow
+- No magic numbers — extract to named constants
+- No deeply nested conditionals (3+ levels) — use early returns or extract functions
+
+## Error Handling
+
+- Handle errors at the boundary, not in every function
+- Never swallow errors silently (`catch (e) {}`)
+- Use typed errors where the language supports it
+- User-facing errors must be human-readable, not stack traces
+
+{{CODE_GUIDELINES_STACK}}

package/templates/ai/workflows/session-checkpoint.md

@@ -43,7 +43,6 @@ Write this at session end. Paste it into the next session's first message.
 ## When to Force a New Session
 
 - Context feels sluggish (repeating yourself, forgetting earlier decisions)
-- 3-5 slices completed since last fresh session
-- Switching between agents (PM → Developer → Tester)
+- 10-15 slices completed since last fresh session
 - After a major refactor or architecture change
 - When the checkpoint itself exceeds 30 lines