Skip to content

DeevsDeevs/agent-system

Repository files navigation

Deevs' Agent System

Deevs' plugin marketplace for Claude Code with workflow chains, terminal control, and expert agents.

Installation

/plugin marketplace add git@github.com:DeevsDeevs/agent-system.git

Codex Setup

Codex loads repo-scoped skills from .codex/skills. This repo includes symlinks to each skill folder, so no extra install step is required.

Usage:

  • Invoke a skill with $skill-name or /skills.
  • For chain operations: $chain-system link <name>, $chain-system load <name>, $chain-system list.
  • For persona workflows: $dev-experts <persona>, $bug-hunters <role>, $research-experts <role>.

If Codex was already running, restart it to reload the skills.

Global (user-scoped) install for any repo:

git clone git@github.com:DeevsDeevs/agent-system.git ~/src/agent-system
mkdir -p ~/.codex/skills
for d in 97-dev anti-ai-slop datetime golang-pro polars-expertise arxiv-search chain-system dev-experts bug-hunters research-experts cost-status; do
  ln -s ~/src/agent-system/$d ~/.codex/skills/$d
done

Installer (interactive by default):

curl -fsSL https://raw.githubusercontent.com/DeevsDeevs/agent-system/main/scripts/install.sh \\
  | bash -s --

Installer flags: --non-interactive, --platform claude|codex|both, --repo-ref <tag|branch|commit>, --uninstall, --skills a,b,c.

Uninstall:

  • Codex: run the installer with --uninstall.
  • Claude Code: use /plugin uninstall <plugin>@deevs-agent-system inside Claude Code.

Install a single skill by URL with $skill-installer.

Plugins

chain-system

Multi-session workflow chains for maintaining context across conversations.

Core: Save conversation state, resume complex projects across sessions without losing technical details.

Commands:

  • /chain-link [name] - Save current work to chain
  • /chain-load [name] - Resume from saved chain
  • /chain-list - List all chains

Use when: Working on multi-session projects, need to pause and resume with full context.

Details: chain-system/README.md

/plugin install chain-system@deevs-agent-system

dev-experts

Critical, opinionated developer personas focused on approach and methodology.

Agents:

/architect - Technical lead and architect

  • Plans features, explores alternatives with trade-off analysis
  • Creates actionable implementation plans
  • Use for: Architecture decisions, feature design

/devops - Production detective

  • Methodical bug hunting and failure investigation
  • Forms hypotheses, tests systematically
  • Use for: Production issues, mysterious bugs, deployment failures

/rust-dev - Rust purist

  • Hunts un-Rusty patterns, ownership issues, safety violations
  • Use for: Rust code review, idiom improvements

/python-dev - Pythonista

  • Type safety, async pitfalls, modern Python (3.10+)
  • Use for: Python code review, modernization

/cpp-dev - C++ performance purist

  • UB, memory bugs, latency killers, lock-free correctness
  • Use for: C++ code review, performance optimization

/reviewer - Grumpy code wizard, 40 years experience

  • Security holes, race conditions, performance sins, edge cases
  • Line-by-line analysis with specific fixes
  • Use for: Pre-merge review, security audit

/tester - Testing specialist

  • Comprehensive, real-world tests (no fake tests)
  • Use for: Writing tests, improving coverage

Details: dev-experts/README.md

/plugin install dev-experts@deevs-agent-system

alpha-squad

Five hypothesis generators attacking from complementary angles. Mechanism over fit. Counterparty over correlation.

Agents:

/fundamentalist - Accounting & value lens

  • Financial statements, earnings quality, capital efficiency
  • Use for: Mispricing from accounting misreads

/vulture - Flows & constraints lens

  • Forced sellers, index reconstitutions, 13F crowding
  • Use for: Constraint-driven alpha, calendar flows

/network-architect - Relationships & propagation lens

  • Customer-supplier networks, lead-lag, contagion paths
  • Use for: Information propagation, network effects

/book-physicist - Microstructure & entropy lens

  • Order book entropy, informativity decomposition, structural models (Kyle, Hawkes)
  • Use for: Microstructure alpha, order flow informativity, entropy regime detection

/causal-detective - Mechanisms & confounding lens

  • Frisch-Waugh-Lovell, Double ML, placebo tests
  • Use for: Proving causal mechanisms, killing confounded signals

Details: alpha-squad/README.md

/plugin install alpha-squad@deevs-agent-system

mft-research-experts

Orchestration, data validation, risk geometry, hypothesis execution, and forensic audit. The infrastructure that makes alpha-squad real - or kills it.

Agents:

/mft-strategist - Obsessive coordinator

  • Orchestrates hypothesis generation and destruction
  • Routes Data Sentinel → Alpha Squad → Factor Geometer → Skeptic
  • Use for: Starting any research, SHIP/KILL/ITERATE decisions

/data-sentinel - Paranoid gatekeeper

  • MUST be invoked FIRST on any data
  • Trusts nothing, asks before filtering
  • Use for: Any data validation

/factor-geometer - Risk architect

  • Factor loadings, covariance estimation, alpha-orthogonal decomposition
  • Use for: Geometry check, factor exposure analysis

/skeptic - Hypothesis executioner

  • Dual causal + statistical validation gauntlet
  • Use for: Full hypothesis validation (Rademacher, walk-forward, placebos)

/forensic-auditor - Post-mortem investigator

  • Traces assumption failures, runs on schedule and crisis
  • Use for: Pipeline review, incident investigation, periodic audit

Details: mft-research-experts/README.md

/plugin install mft-research-experts@deevs-agent-system

bug-hunters

Systematic bug hunting with spec reconstruction, adversarial validation, and confidence-ranked reports.

Agents:

/orchestrator - Central brain (RED)

  • Reconstructs spec from code, spawns hunters, challenges findings
  • Filters false positives via adversarial validation with dev-experts
  • Use for: Starting any bug hunting session

/logic-hunter - Spec detective (ORANGE)

  • Language-agnostic logic bugs, spec-vs-implementation gaps
  • Scan mode (hotspots) → Hunt mode (deep trace)
  • Use for: Algorithm correctness, data flow issues, design intent verification

/cpp-hunter - C++ bug hunter (YELLOW)

  • Memory corruption, UB, concurrency issues
  • Hypothesis-driven, demands proof
  • Use for: Crash debugging, memory bugs, mysterious C++ failures

/python-hunter - Python bug hunter (YELLOW)

  • Async pitfalls, None propagation, type violations
  • Hypothesis-driven, demands proof
  • Use for: Python-specific bugs, type mismatches, async issues

Flow: Spec Reconstruction → Hotspot Scan → Hunter Deployment → Adversarial Challenge → Confidence Scoring → Report

Details: bug-hunters/README.md

/plugin install bug-hunters@deevs-agent-system

cost-status

Status bar showing session cost, monthly cost, total cost, and context window usage.

Display: $0.52/$30.00/$150.00 | 38k/200k (19%)

  • $session/$month/$total - Cost tracking at three levels
  • used/max (%) - Context window usage

Setup: After installation, add to ~/.claude/settings.json:

{
  "statusLine": {
     "type": "command",
     "command": "bash ~/.claude/scripts/show-cost.sh"
   },
}

Find the path with:

find ~/.claude -name "show-cost.sh" -path "*/cost-status/*" 2>/dev/null | head -1

Details: cost-status/README.md

/plugin install cost-status@deevs-agent-system

arxiv-search

Search arXiv preprint repository for academic papers.

Core: Query arXiv for research papers across physics, mathematics, computer science, quantitative biology, finance, and statistics.

Quick start:

# Basic search (auto-selects Python or bash)
arxiv_search "transformer attention mechanism"

# Limit results
arxiv_search "protein folding" --max-papers 5

Use when: Finding preprints, ML/AI papers, mathematical methods, scientific literature before journal publication.

Dependencies: None (bash fallback). For better reliability: uv pip install arxiv

Details: arxiv-search/SKILL.md

/plugin install arxiv-search@deevs-agent-system

Agent Color Scheme

Universal color scheme across all agent plugins:

Color Role Examples
❤️ RED Deciders & Orchestrators architect, mft-strategist, orchestrator, devops
🧡 ORANGE Hybrid (can lead or challenge) logic-hunter
💛 YELLOW Checkers & Validators reviewer, cpp-hunter, python-hunter, skeptic, forensic-auditor
💙 BLUE Builders & Implementers cpp-dev, python-dev, rust-dev, tester
💚 CYAN Researchers & Analysts data-sentinel, fundamentalist, vulture, network-architect, book-physicist, causal-detective, factor-geometer

Credits

Inspired by:

About

Perfect agent system to be 10x Deevs' engineer

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors