264 lines
11 KiB
Markdown
264 lines
11 KiB
Markdown
# Agentic Engineering Skill Kits
|
|
|
|
Production-ready, plug-and-play skill packages from battle-tested agent systems. Each kit includes SKILL.md files, supporting scripts, configurations, and a demo.
|
|
|
|
---
|
|
|
|
## Pricing
|
|
|
|
| Tier | Kit | Price | Skills | What You Get |
|
|
|------|-----|-------|--------|-------------|
|
|
| **Free** | Individual skills | Free | 1 each | Single SKILL.md + basic config |
|
|
| **Starter** | Single kit | **$29-49** | 3-8 | Full kit with scripts + demo |
|
|
| **Pro** | Any 3 kits | **$79-99** | 9-24 | 3 kits bundled |
|
|
| **Enterprise** | Everything | **$199-299** | 40+ | All kits + updates + support |
|
|
|
|
Compare: ClaudeFAST charges $89 for Code Kit, $149 for Complete, $299 for Ultimate. Our kits are priced lower and based on working production systems you can inspect in `tac/`.
|
|
|
|
---
|
|
|
|
## Kit Catalog
|
|
|
|
### 1. Security Foundation Kit — $49
|
|
|
|
**Source**: `bash-damage-from-within/`, `pi-vs-claude-code/damage-control.ts`
|
|
|
|
Protect your agents from catastrophic mistakes. Three security layers in one kit.
|
|
|
|
| Skill | What It Does | Lines |
|
|
|-------|-------------|-------|
|
|
| `l3-blacklist-hook` | Regex blacklist for dangerous bash commands | 45 |
|
|
| `l4-whitelist-hook` | Only N safelisted commands allowed | 55 |
|
|
| `l5-no-bash` | Bash removed entirely, custom tools only | 80 |
|
|
| `damage-control-rules` | YAML rules: zero-access, read-only, no-delete paths | 30 |
|
|
| `agent-sandbox` | E2B sandbox isolation for risky operations | 120 |
|
|
|
|
**Includes**: 5 SKILL.md files + 3 Python hooks + 2 TypeScript extensions + YAML rule files + demo script (6 attack scenarios)
|
|
**Course reference**: M3 Safety & Security
|
|
|
|
### 2. Multi-Agent Orchestration Kit — $49
|
|
|
|
**Source**: `pi-vs-claude-code/`, `lead-agents/`, `ui-agents/`
|
|
|
|
Scale from 1 agent to 20. Orchestration patterns that work.
|
|
|
|
| Skill | What It Does | Lines |
|
|
|-------|-------------|-------|
|
|
| `agent-team` | Dispatcher pattern: orchestrator delegates to specialists | 150 |
|
|
| `agent-chain` | Pipeline pattern: YAML-defined sequential workflows | 120 |
|
|
| `p2p-coms` | Peer-to-peer agent communication (Unix sockets) | 200 |
|
|
| `p2p-coms-net` | Networked P2P agent communication (HTTP/SSE) | 180 |
|
|
| `tilldone` | Task discipline system with live progress tracking | 100 |
|
|
| `purpose-gate` | Session intent declaration before work starts | 60 |
|
|
| `mental-model` | Agent-owned expertise files that compound across sessions | 75 |
|
|
| `domain-lock` | Per-agent file system permissions (read/write/delete scope) | 65 |
|
|
|
|
**Includes**: 8 SKILL.md files + 4 TypeScript extensions + 3 YAML config templates + team config examples + demo
|
|
**Course reference**: M4 Multi-Agent Orchestration
|
|
|
|
### 3. Verifier Pro Kit — $39
|
|
|
|
**Source**: `the-verifier-agent/`
|
|
|
|
Catch agent mistakes before they hit production. Read-only verification that runs alongside your builder agent.
|
|
|
|
| Skill | What It Does | Lines |
|
|
|-------|-------------|-------|
|
|
| `verifier-builder` | Two-agent observer: builder + verifier over Unix socket | 250 |
|
|
| `confidence-ladder` | PERFECT → VERIFIED → PARTIAL → FEEDBACK → FAILED grading | 40 |
|
|
| `claim-decomposition` | Break every claim into atomic verifiable units | 80 |
|
|
| `read-only-surface` | Restrict verifier to read/grep/find/ls only | 30 |
|
|
| `escalation-handler` | Max 3 correction loops, then escalate to human | 50 |
|
|
|
|
**Includes**: 5 SKILL.md files + TypeScript extension + Python verifier script + status bar UI + demo
|
|
**Course reference**: M3 (Verifier Pattern)
|
|
|
|
### 4. Task Discipline Kit — $29
|
|
|
|
**Source**: `pi-vs-claude-code/extensions/tilldone.ts`
|
|
|
|
Stop vibe coding. Start engineering with structured task management.
|
|
|
|
| Skill | What It Does | Lines |
|
|
|-------|-------------|-------|
|
|
| `tilldone-core` | Task list gating: no work without defined tasks | 100 |
|
|
| `tilldone-progress` | Live task progress in footer widget | 60 |
|
|
| `tilldone-nudge` | Agent prompted to continue if tasks remain | 30 |
|
|
| `purpose-gate` | Declare session intent before first prompt | 60 |
|
|
|
|
**Includes**: 4 SKILL.md files + TypeScript extension + JSON config + demo
|
|
**Course reference**: M4 (TillDone)
|
|
|
|
### 5. Autoresearch Kit — $39
|
|
|
|
**Source**: `mythos-learnings.md`, `brand-monitor/autoresearch`
|
|
|
|
Agents that improve themselves. Run experiments, measure results, keep what works.
|
|
|
|
| Skill | What It Does | Lines |
|
|
|-------|-------------|-------|
|
|
| `experiment-loop` | Run → Measure → Log → Decide (keep/discard) | 120 |
|
|
| `integrity-guard` | Code hashing to detect grinding | 30 |
|
|
| `median-over-best` | Compare against median, not best (prevents noise-chasing) | 20 |
|
|
| `reward-hack-defense` | Detect when model moves computation outside timing | 40 |
|
|
| `log-to-jsonl` | Structured experiment logging to JSONL | 35 |
|
|
|
|
**Includes**: 5 SKILL.md files + Python experiment harness + JSONL log viewer + demo with sample data
|
|
**Course reference**: M7 (Autoresearch)
|
|
|
|
### 6. Observability Kit — $29
|
|
|
|
**Source**: `hooks-multi-agent-observability/`, `claude-code-hooks-mastery/`
|
|
|
|
See what your agents are actually doing. Every tool call, every LLM completion, every cost.
|
|
|
|
| Skill | What It Does | Lines |
|
|
|-------|-------------|-------|
|
|
| `tool-call-tracer` | Log every tool call to SQLite with params + result | 100 |
|
|
| `cost-tracker` | Track per-session and per-agent costs in real-time | 60 |
|
|
| `session-replay` | Replay agent sessions for debugging | 80 |
|
|
| `loop-detector` | Alert when same tool called 5+ times with same params | 25 |
|
|
| `observability-dashboard` | Vue dashboard showing live agent activity | 200 |
|
|
|
|
**Includes**: 5 SKILL.md files + Python SQLite logger + TypeScript extension + Vue dashboard + demo
|
|
**Course reference**: M5 (Observability)
|
|
|
|
### 7. CEO Board Kit — $49
|
|
|
|
**Source**: `ceo-agents/`
|
|
|
|
8 specialist agents for strategic decision-making. Adversarial debate, fact verification, ranked recommendations.
|
|
|
|
| Skill | What It Does | Lines |
|
|
|-------|-------------|-------|
|
|
| `ceo-orchestrator` | Frames decisions, opens board meetings, synthesizes | 150 |
|
|
| `revenue-agent` | 30-90 day cash-focused perspective | 80 |
|
|
| `compounder-agent` | 6-24 month trust-focused perspective | 80 |
|
|
| `contrarian-agent` | Attacks assumptions, 3x weight on dissent | 80 |
|
|
| `technical-architect` | System durability and technical debt | 80 |
|
|
| `product-strategist` | Problem selection and prioritization | 80 |
|
|
| `customer-oracle` | User behavior and market signals | 80 |
|
|
| `market-strategist` | Competitive positioning and timing | 80 |
|
|
| `moonshot-agent` | Asymmetric upside and long bets | 80 |
|
|
| `verifier-agent` | Fact-check: 2+ sources per claim | 100 |
|
|
| `tracker-agent` | Decision logging, quarterly review scheduling | 60 |
|
|
|
|
**Includes**: 11 SKILL.md files + TypeScript extension + brief templates + SVG diagram generator + demo
|
|
**Course reference**: M4 (CEO Board), M6 (Evaluation)
|
|
|
|
### 8. Plan F3 — Free individual skill
|
|
|
|
**Source**: [`disler/planf3`](https://github.com/disler/planf3), installed at `skills/individual/planf3/`
|
|
|
|
HTML-first implementation planning for Mythos/Fable-class workflows. Produces browser-openable plans in `specs/`, with metadata, phase checklists, validation loops, amendments, and optional image generation.
|
|
|
|
| Skill | What It Does | Lines |
|
|
|-------|-------------|-------|
|
|
| `planf3` | Create/update/build detailed HTML-first engineering plans | 250+ |
|
|
|
|
**Includes**: `SKILL.md`, workflow docs, and image-generation helper scripts.
|
|
|
|
---
|
|
|
|
## Enterprise Bundle — $199 (save $124)
|
|
|
|
All 7 kits (40+ skills) + everything below:
|
|
|
|
| Bonus | Value |
|
|
|-------|-------|
|
|
| Priority updates (new model releases, framework changes) | — |
|
|
| Private Discord channel for setup support | — |
|
|
| Custom skill development (2 skills of your choice) | $200 value |
|
|
| 1-hour onboarding call with the engineer | $150 value |
|
|
| Early access to new kits before public release | — |
|
|
|
|
---
|
|
|
|
## Individual Skills (à la carte)
|
|
|
|
Available individually for **free** (basic) or **$9** (full with scripts + support):
|
|
|
|
| Skill | Basic (free) | Full ($9) |
|
|
|-------|-------------|-----------|
|
|
| `damage-control-rules` | YAML rules only | + TypeScript extension + 3 access levels |
|
|
| `tilldone-core` | SKILL.md only | + Extension + progress widget |
|
|
| `verifier-confidence` | SKILL.md only | + Python verifier + status bar |
|
|
| `experiment-loop` | SKILL.md only | + Python harness + JSONL viewer |
|
|
| `tool-call-tracer` | SKILL.md only | + SQLite logger + queries |
|
|
| `p2p-coms-basic` | SKILL.md only | + TypeScript extension + demo |
|
|
| `agent-chain-basic` | SKILL.md only | + YAML templates + examples |
|
|
| `mental-model` | SKILL.md only | + Self-improve script |
|
|
| `domain-lock` | SKILL.md only | + YAML config generator |
|
|
| `purpose-gate` | SKILL.md only | + Extension + config |
|
|
|
|
---
|
|
|
|
## How to Use a Kit
|
|
|
|
```bash
|
|
# Each kit is self-contained:
|
|
kits/security-foundation/
|
|
├── SKILL.md # Master skill declaration
|
|
├── skills/
|
|
│ ├── l3-blacklist/
|
|
│ │ ├── SKILL.md
|
|
│ │ └── blacklist.py
|
|
│ ├── l4-whitelist/
|
|
│ │ ├── SKILL.md
|
|
│ │ └── whitelist.py
|
|
│ └── damage-control/
|
|
│ ├── SKILL.md
|
|
│ └── rules.yaml
|
|
├── extensions/
|
|
│ └── damage-control.ts
|
|
├── demo/
|
|
│ └── run-demo.sh
|
|
└── README.md
|
|
|
|
# Install:
|
|
cp -r skills/ ~/.pi/skills/ # Pi Agent
|
|
cp -r skills/ ~/.claude/skills/ # Claude Code
|
|
# Or use the install script:
|
|
bash install.sh
|
|
```
|
|
|
|
---
|
|
|
|
## Kit Comparison: Our Kits vs ClaudeFAST
|
|
|
|
| Dimension | ClaudeFAST | Our Kits |
|
|
|-----------|-----------|----------|
|
|
| **Price (single)** | $89-99 | $29-49 |
|
|
| **Price (bundle)** | $149-299 | $79-199 |
|
|
| **Skills count** | 280 (across all kits) | 40+ (across all kits) |
|
|
| **Open source** | No | Yes (MIT) |
|
|
| **Source code** | Closed | Full `tac/` repo with production systems |
|
|
| **Course integration** | Separate blog | Designed alongside M1-M8 course |
|
|
| **Original research** | No | mythos-learnings, bash-damage, verifier |
|
|
| **Format** | SKILL.md + downloads | SKILL.md + TypeScript/Python + demo |
|
|
|
|
### Why Our Kits Are Cheaper
|
|
|
|
1. Our source material is already in the `tac/` repo — no additional development cost
|
|
2. We're building on open-source tools (Pi, Claude Code, OpenCode)
|
|
3. The course and kits are designed together — no separate production
|
|
|
|
---
|
|
|
|
## Next Skills to Build (Roadmap)
|
|
|
|
| Kit | Status | Est. Release |
|
|
|-----|--------|-------------|
|
|
| Security Foundation | Ready | Now |
|
|
| Multi-Agent Orchestration | Ready | Now |
|
|
| Verifier Pro | Ready | Now |
|
|
| Task Discipline | Ready | Now |
|
|
| Autoresearch | Ready | Now |
|
|
| Observability | Ready | Now |
|
|
| CEO Board | Ready | Now |
|
|
| Enterprise Bundle | Ready | Now |
|
|
| **Prompt Injection Defense** | Planned | Q3 2026 |
|
|
| **MCP Server Toolkit** | Planned | Q3 2026 |
|
|
| **Agent Testing Suite** | Planned | Q4 2026 |
|
|
| **Deployment Pipeline** | Planned | Q4 2026 |
|