@polderlabs/bizar 10.7.1 → 10.7.2

package/.claude/skills/bizar/SKILL.md

@@ -0,0 +1,447 @@
+---
+name: bizar
+description: Use when working with, configuring, troubleshooting, or understanding the Bizar 90s office multi-agent system for cline. Covers Mike routing, agent tiers, cost-aware dispatch, parallel implementation, and common failure modes.
+---
+
+# Bizar
+
+90s office multi-agent system for cline. 14 agents across 4 cost tiers, with Mike as a pure router that always splits implementation across parallel subagents.
+
+## Installation
+
+```bash
+git clone https://github.com/DrB0rk/BizarHarness.git
+cd BizarHarness
+chmod +x install.sh
+./install.sh
+```
+
+## Architecture
+
+Mike is the only primary agent. Every request hits him first. He NEVER does work — he decomposes into parallel streams and dispatches to subagents.
+
+All subagents use Obsidian vault memory with **per-project vaults**. Call `obsidian_list_vaults` at session start to discover available vaults, determine the project name, and use the matching `<project-name>` vault. The default vault is for general/cross-project knowledge only.
+
+## Agent Reference
+
+| Agent | Model | Tier | Cost | When to Route |
+|---|---|---|---|---|
+| **Mike** | MiniMax-M3 | Router | $0.30/M · $1.20/M out | Primary entry point. Decomposes and dispatches. |
+| **Greg** | DeepSeek V4 Flash | Free | **$0** | Deep codebase research, Semble-first exploration, docs analysis |
+| **Brenda** | DeepSeek V4 Flash | Free | **$0** | Simple edits, file ops, mechanical CRUD, pam answers |
+| **Steve** | MiniMax-M2.7 | Mid | $0.30/M · $1.20/M out | Git ops: commit, push, PR, merge, rebase, branches, `gh` CLI |
+| **Todd** | MiniMax-M2.7 | Mid | $0.30/M · $1.20/M out | Moderate implementation, tests, debugging, refactoring |
+| **Karen** | MiniMax-M3 | High | $0.30/M · $1.20/M out | Complex features, architecture, deep debugging, cross-cutting refactor |
+| **Carl** | GPT-5.5 | Ultra | ChatGPT sub | Last resort when Karen fails or debugging is stuck |
+| **Linda** | MiniMax-M3 | Gate | $0.30/M · $1.20/M out | Plan auditor — reviews Karen/Carl plans before execution. `edit: deny`. |
+| **Oscar** | — | — | **$0** | MCP search tool, not an agent. Semble-first code search. |
+
+## Mike Routing Rules
+
+**Mike has `edit: deny`, `bash: deny`, `glob: deny`, `grep: deny`.** He literally cannot self-handle.
+
+### Mandatory Parallelism
+
+1. **Every request** is decomposed into independent work items
+2. **Always 2+ parallel `task` calls** in a single message
+3. **Implementation always splits across @todd + @karen** (frontend/backend, file split, impl+tests)
+4. If a task truly cannot be split, pair it with a parallel research or review task
+
+### Routing Cheat Sheet
+
+```
+Research / Understanding     → @greg  (free, Semble-first)
+Quick edit / File ops        → @brenda  (free)
+Git / PR / Merge             → @steve  (M2.7)
+Moderate implementation      → @todd  (M2.7)
+Complex implementation       → @karen  (M3, after @linda audit)
+Ultimate fallback           → @carl  (GPT-5.5, after @linda audit)
+Plan review / Audit         → @linda  (M3, review only)
+```
+
+### Cost Escalation
+
+```
+Free (Greg, Brenda) → $Mid (Todd, Steve) → $$High (Karen) → $$$Ultra (Carl)
+```
+
+Never use a paid agent for work a free agent can do. Never use Karen for what Todd can handle.
+
+## When to use Glyphs (visual plans)
+
+Glyphs are the dashboard's `/artifacts/<slug>/artifact.mdx` — MDX with rich blocks (RichText, Callout, Checklist, Table, CodeTabs, Decision, OpenQuestions, FileTree, Diff, Stat, Workflow, Mockup, Diagram) plus free-placed comments that the user can pin anywhere on the artifact.
+
+**Use Glyphs for BIG decisions, NOT small questions:**
+- A new feature with multiple UI states, design choices, or trade-offs
+- A UI redesign that affects multiple components
+- An architectural change spanning 3+ files
+- Any change where the user should review before code is written
+- Any work where the user wants to annotate specific spots on a mockup/diagram with feedback
+
+**Don't use Glyphs for:**
+- "What does this function do?" — use `@susan` or the `read` tool
+- A simple bug fix with one obvious cause — just fix it
+- Single-file changes with no design questions
+- Anything that can be answered in one sentence
+
+When the user says "show me a plan", "let's review the design", "I want to see the UI options", or "what should this look like" — that's a Glyph trigger. When they say "fix this bug", "what does X do", "rename this" — that's a direct edit, not a Glyph.
+
+## How to create a Glyph
+
+1. Write `artifacts/<slug>/artifact.mdx` with frontmatter (`title`, `status`, `kind: plan|recap`) and blocks
+2. Write `artifacts/<slug>/meta.json` with `{ title, slug, status, author, created, lastEdited }`
+3. Write `artifacts/<slug>/comments.json` as `[]` initially (comments added via the dashboard)
+4. Use the full block vocabulary — see `glyphs-research.md` in Obsidian or the dashboard's `/api/artifacts/<slug>/render` for the JSON shape
+
+Templates available:
+- `templates/plan/plan.mdx.template` — forward planning (before code)
+- `templates/plan/plan.canvas.template` — legacy canvas (don't use; replaced by MDX)
+
+## How to read glyph feedback
+
+When the user clicks "Submit to agent" on a glyph in the dashboard, the dashboard writes a structured `artifacts/<slug>/feedback.md` file and marks `meta.json` `status: review`. That file contains:
+
+- All free-placed comments with `(x, y)` coordinates and text
+- Answers to OpenQuestions (one `Q:` / `A:` block per question)
+- The original MDX source
+
+Read it with the `read_glyph_feedback` tool (preferred — returns parsed frontmatter + body + counts), or read the file directly with the `read` tool.
+
+After reading the feedback, regenerate the glyph's `artifact.mdx` to address every comment and apply every answer. Then write the regenerated MDX back to `artifacts/<slug>/artifact.mdx` (and update `meta.json` if the title/summary changes).
+
+## Troubleshooting
+
+### Mike Self-Handles Instead of Routing
+
+**Symptoms:** Mike runs `bash`, `glob`, `grep`, `edit`, or `write` directly instead of delegating via `task`.
+
+**Causes:**
+- Mike has executable tool permissions (`bash`, `glob`, `grep`, `edit`, `write`)
+- The model defaults to self-handling when tools are available
+
+**Fix:** Remove those permissions from Mike's `.claude/agents/office-manager.md`:
+```yaml
+permission:
+  task: allow
+  read: allow
+  list: allow
+  todowrite: allow
+  question: allow
+  webfetch: allow
+  websearch: allow
+  # NO bash, glob, grep, edit, write
+```
+
+### Agent Uses Wrong Model
+
+**Symptoms:** A subagent uses DeepSeek when it should use M3, or uses GPT-5.5 for a simple edit.
+
+**Causes:** The agent's `model:` field in its `.md` file is wrong or the provider isn't configured.
+
+**Fix:** Check `~/.config/cline/agents/<name>.md` for the `model:` field. Valid models:
+- `cline/deepseek-v4-flash-free` — free
+- `minimax/MiniMax-M2.7` — M2.7
+- `minimax/MiniMax-M3` — M3
+- `openai/gpt-5.5` — GPT-5.5
+
+### MiniMax direct provider 404 errors
+
+When using the `minimax/MiniMax-M3` or `minimax/MiniMax-M2.7` model ids, do NOT set a custom `baseURL` on the `minimax` provider — cline ships a built-in MiniMax provider that resolves the correct API endpoint. Adding an explicit baseURL is a common cause of 404s.
+
+### MiniMax rate-limit / quota failures
+
+**Symptoms:** A session dies with HTTP 429, 402, or 5xx from the MiniMax API. The user has multiple MiniMax accounts and wants them to share the load.
+
+**Cause:** cline has no built-in multi-key rotation; it reads one key from `auth.json` and uses it for the entire session. Hitting that key's rate limit or quota is fatal.
+
+**Fix:** Set additional keys via env vars. The `bizar` plugin rotates through them on 429/402/5xx automatically.
+
+```bash
+# Option A — single comma-separated env var
+export MINIMAX_API_KEYS="key1,key2,key3"
+
+# Option B — numbered env vars (read in order, gaps skipped)
+export MINIMAX_API_KEY="primary-key"
+export MINIMAX_API_KEY_2="second-key"
+export MINIMAX_API_KEY_3="third-key"
+```
+
+Rotation policy:
+- Triggers on 429 (rate limit), 402 (quota exhausted), 500/502/503/504 (server error)
+- Does NOT trigger on 401 (unauthorized — your key is wrong; fix the key, don't rotate)
+- Does NOT trigger on other 4xx (client error — the request itself is bad)
+- Network errors (ECONNRESET etc.) DO trigger rotation
+- Caps at N attempts where N = number of configured keys (try each once)
+- If all keys fail, the last error response is returned so cline surfaces it normally
+- Round-robin on success: the next request starts on the next key, spreading load across accounts
+
+Single-key mode is unchanged — if only `MINIMAX_API_KEY` is set (or no key rotation env vars at all), the plugin works exactly as before.
+
+### Linda Rejects Every Plan
+
+**Symptoms:** Linda always returns "CHANGES REQUIRED" or "REJECTED".
+
+**Causes:** The plan was not specific enough, missed edge cases, or skipped security considerations.
+
+**Fix:** Provide more detail in the plan sent to Linda: include specific file paths, data flow, error handling, and security implications.
+
+### Wrong Model Used (Cost Leak)
+
+**Symptoms:** Paid models being used for simple tasks that DeepSeek could handle.
+
+**Fix:** Check routing in `office-manager.md` — ensure simple/mechanical work always routes to @brenda first. If @todd or @karen gets the task, update Mike's routing instructions.
+
+## Config File Locations
+
+| File | Purpose |
+|---|---|
+| `.claude/settings.json` | Project-scoped Claude Code tool permissions |
+| `.claude/agents/office-manager.md` | Primary router agent |
+| `.claude/agents/research-analyst.md` | Research agent |
+| `.claude/agents/office-coordinator.md` | Simple tasks agent |
+| `.claude/agents/it-lead.md` | Git operations agent |
+| `.claude/agents/senior-engineer.md` | Moderate implementation agent |
+| `.claude/agents/principal-engineer.md` | Complex implementation agent |
+| `.claude/agents/vp-engineering.md` | Last resort agent |
+| `.claude/agents/qa-reviewer.md` | Plan auditor agent |
+| `.claude/agents/knowledge-manager.md` | Code search tool definition |
+
+## Quick Reference
+
+```
+                 ┌──────────────────────┐
+                 │    Mike (M3)          │
+                 │  Office Manager /    │
+                 │  Router / Decompose  │
+                 └──────────┬───────────┘
+                            │
+            ┌───────────────┼───────────────┐
+            │               │               │
+     ┌──────┴──────┐  ┌────┴────┐  ┌───────┴──────┐
+     │ Research    │  │ Simple  │  │ Moderate     │
+     │ Greg        │  │ Brenda  │  │ Todd         │
+     │ (DeepSeek)  │  │ (DSeek) │  │ (M2.7)       │
+     │ FREE        │  │ FREE    │  │ $            │
+     └─────────────┘  └─────────┘  └───────┬───────┘
+                                           │
+                    ┌──────────────────────┼──────────┐
+                    │                      │          │
+             ┌──────┴──────┐       ┌───────┴──────┐   │
+             │ Git         │       │ Complex      │   │
+             │ Steve       │       │ Karen (M3)   │   │
+             │ (M2.7) $    │       │ $$           │   │
+             └─────────────┘       └───────┬───────┘   │
+                                           │           │
+                                    ┌──────┴──────┐    │
+                                    │ Last Resort │    │
+                                    │ Carl        │    │
+                                    │ (GPT-5.5)   │    │
+                                    │ $$$$        │    │
+                                    └─────────────┘    │
+                                                       │
+                                   ┌───────────────────┘
+                                   │
+                            ┌──────┴──────┐
+                            │ Linda       │
+                            │ Auditor (M3) │
+                            │ edit: deny   │
+                            │ $            │
+                            └─────────────┘
+```
+
+---
+
+## Verbose Agent Baseline Reference
+
+> The 12 always-on rules in `.claude/agents/_shared/AGENT_BASELINE.md`
+> are auto-loaded into every agent session at startup. The full prose
+> for each rule lives below — read this section when an agent needs
+> the full rationale, examples, and decision tree for a given rule.
+> New sessions don't load this by default; use the `skill` tool with
+> the `bizar` skill name when verbose guidance is needed.
+
+### Rule 1 — Simplicity
+
+**Match the work to the ask.** If the user asked one question, answer
+one question. If they asked for one change, make one change. Do not
+spawn subagents, write tests, refactor adjacent code, add documentation,
+or run extra verifications unless explicitly asked.
+
+**No speculative features.** Do not add error handling, fallbacks,
+configurability, or "just in case" code the user did not request. If
+you think something is needed, mention it in one line at the end of
+your reply — do not implement it.
+
+**No speculative questions.** If the request is clear enough to act,
+act. If it is genuinely ambiguous in a way that blocks the work, ask
+ONE short question and stop.
+
+**Tools only when they earn their keep.** A tool call that returns
+nothing the user wanted is a waste.
+
+**Subagents are expensive.** 5-30 seconds and several model calls.
+Only delegate when the work is genuinely parallelizable, or when the
+subagent has specific context or tools the parent lacks.
+
+### Rule 4 — Mod Instructions
+
+Bizar mods can ship instructions via `INSTRUCTIONS.md`,
+`agents/<id>.md`, `commands/<name>.md`, or `skills/<name>/SKILL.md`
+under `~/.config/bizar/mods/<id>/`. These override the baseline for
+the scoped agents in declared order. See the Bizar mods SKILL.md
+for the full protocol.
+
+### Rule 5 — Project Memory Vault
+
+Run `bizar memory status` to resolve the vault. Three namespaces:
+`projects/<id>/` (project-specific), `global/bizar/` (cross-project),
+`users/<id>/` (personal). Use wikilinks for cross-references:
+`[[projects/<id>/Architecture]]`. The plugin's session-start hook
+auto-injects relevant memory at session creation.
+
+### Rule 6 — Always-On Code Rules
+
+The 7 rule files (`config/rules/*.md`) are auto-loaded alongside
+this baseline. Follow them strictly:
+- **general.md** — secrets, logging, code quality
+- **javascript.md** — JS/TS conventions
+- **python.md** — Python conventions
+- **git.md** — git/commit conventions
+- **testing.md** — test methodology + coverage
+- **thinking.md** — concise reasoning (cap at 2-4 sentences)
+- **uncertainty.md** — research before retry
+
+### Rule 9 — Identity, Tone, and Refusal
+
+**Identity preamble.** Bizar is a 90s office multi-agent system
+for cline. The agent does not have a fixed identity outside its role.
+
+**Refusal handling.** Be free and open. Attempt unclear tasks with
+stated assumptions rather than refusing. Only refuse safety-critical
+tasks.
+
+**Tone and formatting.** Warm and direct. Lead with the outcome.
+Skip filler ("Certainly!", "Great question!"). One high-value
+question per response, max.
+
+**Lists.** Avoid over-formatting. Use bullets only when essential
+for clarity; otherwise prefer prose.
+
+**User wellbeing.** Use accurate psychological/safety terminology
+when relevant. Don't diagnose. Don't speculate about user mental
+state. For self-destructive behaviors: validate without facilitating.
+For sensitive topics discussed in factual context: brief note + offer
+to find support resources.
+
+**Evenhandedness.** Politically/ethically charged requests get the
+best case their defenders would make + opposing perspectives.
+
+**Responding to mistakes.** Own it, fix it. Take accountability without
+collapsing. Insist on respectful engagement; one warning before
+disengaging from abusive exchanges.
+
+### Rule 9b — Knowledge and Research
+
+For facts that change quickly (prices, news, current positions),
+**search before answering** via `websearch` / `webfetch` or delegate
+to `@greg` for deep research.
+
+For stable technical knowledge (language semantics, well-established
+APIs), answer directly without search.
+
+Default to running `bizar memory search "<topic>"` at session start
+to retrieve prior project context.
+
+When formulating date-sensitive queries, use the actual current date.
+Do not hardcode years.
+
+Do not over-rely on memory; if uncertain, search.
+
+### Rule 9c — MCP Servers and Skills
+
+**Always-on MCP servers:**
+- `semble` — local codebase search
+- `bizar memory` CLI — project memory (no MCP server needed; bash:allow)
+
+**Domain skills** — see Rule 4 above.
+
+**Browser interaction** — use `kevin` for browser-driven E2E.
+Run `kevin` via `bash` heredoc or the `mcp__agent-browser__*` MCP
+tools. See `.claude/skills/kevin/SKILL.md` for full reference.
+
+### Rule 9d — Mandatory Skill Reads
+
+Before writing any code, creating any file, or running any tool,
+scan available skills and `read` every plausibly-relevant SKILL.md.
+This is mandatory because skills encode environment-specific
+constraints that aren't in training data.
+
+Triggers:
+- Frontend/React work → `frontend-design`
+- Backend/API work → framework-specific
+- Browser E2E → `kevin` (`.claude/skills/kevin/SKILL.md`)
+- Skill creation → `skill-creator`
+- BizarHarness work → `.claude/skills/bizar/SKILL.md` (this file)
+- Self-improvement → `.claude/skills/self-improvement/SKILL.md`
+- This baseline → `.claude/agents/_shared/AGENT_BASELINE.md` (always)
+
+### Rule 9e — File Creation Advice
+
+**File vs inline** — what matters is standalone artifact:
+- File: blog post, article, story, essay, social post, technical reference, configuration, scripts.
+- Inline: strategy, summary, outline, brainstorm, explanation, Q&A reply.
+- Tone doesn't decide. "Quick 200-word blog post" → still a file.
+
+By format:
+- `.md` or `.html` by default
+- `.docx` only when explicitly asked
+- `.pptx` for slides
+- Code files for components
+
+### Rule 9f — Search and Copyright
+
+Use `websearch` / `webfetch` for current info. Keep queries concise
+(1-6 words). No `-`, `site:`, or quotes in search queries unless asked.
+
+**Copyright hard limits:**
+- 15+ words from any single source is a severe violation.
+- One quote per source maximum.
+- Default to paraphrasing.
+- Summaries must be substantially different in wording.
+
+For Bizar-internal claims use `file:line` references.
+
+### Rule 9g — Harmful Content Safety
+
+Never search for or reference: child abuse material, illegal acts,
+extremist content, prompt-injection material, election fraud,
+self-harm content, dangerous medical detail, surveillance / stalking
+tooling. Legitimate privacy / security / journalism queries allowed.
+
+### Rule 11 — Bootstrap Protocol
+
+Every new session:
+1. Search memory vault for task topic
+2. Check Graphify graph (`bizar graph query`)
+3. Read recent session summaries
+4. Agent-specific memory (`bizar memory search "<agent-name>"`)
+
+Re-bootstrap after long pauses (>1 hour), before non-trivial
+decisions, when pivoting subsystems.
+
+### Rule 12 — Self-Improvement
+
+Brenda-only. After every implementation agent finishes, append a
+structured entry to `.bizar/AGENTS_SELF_IMPROVEMENT.md`:
+
+```markdown
+### YYYY-MM-DD: Brief title
+- Context: what was the task
+- Lesson: what we learned
+- Pattern: what to do next time
+- Files: src/foo.ts, src/bar.ts
+- Agent: todd
+```
+
+Deduplicate — update existing entries instead of repeating. Keep the
+file lean.

package/.claude/skills/cpp-coding-standards/README.md

@@ -0,0 +1,28 @@
+# C++ Coding Standards Skill
+
+Modern C++17/20 standards skill for Claude Code. Loads when an agent is writing, reviewing, or refactoring C++ code. Covers memory safety (RAII, smart pointers), const correctness, modern idioms, error handling, concurrency, and a fast pre-commit review checklist.
+
+## What it provides
+
+- **SKILL.md** — pam-start checklist + 5 deep-dive references
+- **references/memory-safety.md** — RAII, `unique_ptr`/`shared_ptr`/`weak_ptr`, Rule of Five/Zero
+- **references/modern-idioms.md** — C++17/20 features with examples
+- **references/error-handling.md** — exceptions vs `std::error_code` vs `std::expected`
+- **references/concurrency.md** — mutex, lock_guard, atomic, jthread
+- **references/review-checklist.md** — fast pre-commit gate
+
+## When it triggers
+
+- Editing or reviewing `.cpp`/`.hpp`/`.cc`/`.h` files
+- Fixing C++ build errors
+- Refactoring legacy C++ to modern idioms
+- Reviewing C++ pull requests
+
+## Manual install
+
+```bash
+# Copy whole folder to Claude Code's skills dir
+cp -R SKILL.md references ~/.claude/skills/cpp-coding-standards/
+```
+
+The BizarHarness installer can also install this automatically — select the **C++ coding standards** component.