{
  "markdown": "# Deevs' Agent System\n\nDeevs' plugin marketplace for Claude Code with workflow chains, terminal control, and expert agents.\n\n## Installation\n\n```bash\n/plugin marketplace add git@github.com:DeevsDeevs/agent-system.git\n```\n\n## Codex Setup\n\nCodex loads repo-scoped skills from `.agents/skills`. This repo includes `.agents/skills` symlinks to each skill folder.\nFor backward compatibility, `.codex/skills` symlinks are also kept.\n\nCodex multi-agent role config for all personas is included in:\n- `.codex/config.toml`\n- `.codex/agents/*.toml`\n\nUsage:\n- Invoke a skill with `$skill-name` or `/skills`.\n- For chain operations: `$chain-system link <name>`, `$chain-system load <name>`, `$chain-system list`.\n- For persona workflows: `$dev-experts <persona>`, `$bug-hunters <role>`, `$alpha-squad <role>`, `$mft-research-experts <role>`.\n\nIf Codex was already running, restart it to reload the skills.\n\nGlobal (user-scoped) install for any repo:\n\n```bash\ncurl -fsSL https://raw.githubusercontent.com/DeevsDeevs/agent-system/main/scripts/install.sh \\\n  | bash -s -- --non-interactive --platform codex\n```\n\nInstall selected skills only:\n\n```bash\ncurl -fsSL https://raw.githubusercontent.com/DeevsDeevs/agent-system/main/scripts/install.sh \\\n  | bash -s -- --non-interactive --platform codex --skills dev-experts,bug-hunters\n```\n\nInstaller (interactive by default):\n\n```bash\ncurl -fsSL https://raw.githubusercontent.com/DeevsDeevs/agent-system/main/scripts/install.sh \\\\\n  | bash -s --\n```\n\nInstaller flags: `--non-interactive`, `--platform claude|codex|both`, `--repo-ref <tag|branch|commit>`, `--uninstall`, `--skills a,b,c`.\n\nUninstall:\n- Codex: run the installer with `--uninstall`.\n- Claude Code: use `/plugin uninstall <plugin>@deevs-agent-system` inside Claude Code.\n\nInstall a single skill by URL with `$skill-installer`.\n\n## Plugins\n\n### chain-system\n\nMulti-session workflow chains for maintaining context across conversations.\n\n**Core**: Save conversation state, resume complex projects across sessions without losing technical details.\n\n**Commands**:\n- `/chain-link [name]` - Save current work to chain\n- `/chain-load [name]` - Resume from saved chain\n- `/chain-list` - List all chains\n\n**Use when**: Working on multi-session projects, need to pause and resume with full context.\n\n**Details**: [chain-system/README.md](chain-system/README.md)\n\n```bash\n/plugin install chain-system@deevs-agent-system\n```\n\n---\n\n### dev-experts\n\nCritical, opinionated developer personas focused on approach and methodology.\n\n**Agents**:\n\n**`/architect`** - Technical lead and architect\n- Plans features, explores alternatives with trade-off analysis\n- Creates actionable implementation plans\n- Use for: Architecture decisions, feature design\n\n**`/devops`** - Production detective\n- Methodical bug hunting and failure investigation\n- Forms hypotheses, tests systematically\n- Use for: Production issues, mysterious bugs, deployment failures\n\n**`/rust-dev`** - Rust purist\n- Hunts un-Rusty patterns, ownership issues, safety violations\n- Use for: Rust code review, idiom improvements\n\n**`/python-dev`** - Pythonista\n- Type safety, async pitfalls, modern Python (3.10+)\n- Use for: Python code review, modernization\n\n**`/cpp-dev`** - C++ performance purist\n- UB, memory bugs, latency killers, lock-free correctness\n- Use for: C++ code review, performance optimization\n\n**`/reviewer`** - Grumpy code wizard, 40 years experience\n- Security holes, race conditions, performance sins, edge cases\n- Line-by-line analysis with specific fixes\n- Use for: Pre-merge review, security audit\n\n**`/tester`** - Testing specialist\n- Comprehensive, real-world tests (no fake tests)\n- Use for: Writing tests, improving coverage\n\n**Details**: [dev-experts/README.md](dev-experts/README.md)\n\n```bash\n/plugin install dev-experts@deevs-agent-system\n```\n\n---\n\n### alpha-squad\n\nFive hypothesis generators attacking from complementary angles. Mechanism over fit. Counterparty over correlation.\n\n**Agents**:\n\n**`/fundamentalist`** - Accounting & value lens\n- Financial statements, earnings quality, capital efficiency\n- Use for: Mispricing from accounting misreads\n\n**`/vulture`** - Flows & constraints lens\n- Forced sellers, index reconstitutions, 13F crowding\n- Use for: Constraint-driven alpha, calendar flows\n\n**`/network-architect`** - Relationships & propagation lens\n- Customer-supplier networks, lead-lag, contagion paths\n- Use for: Information propagation, network effects\n\n**`/book-physicist`** - Microstructure & entropy lens\n- Order book entropy, informativity decomposition, structural models (Kyle, Hawkes)\n- Use for: Microstructure alpha, order flow informativity, entropy regime detection\n\n**`/causal-detective`** - Mechanisms & confounding lens\n- Frisch-Waugh-Lovell, Double ML, placebo tests\n- Use for: Proving causal mechanisms, killing confounded signals\n\n**Details**: [alpha-squad/README.md](alpha-squad/README.md)\n\n```bash\n/plugin install alpha-squad@deevs-agent-system\n```\n\n---\n\n### mft-research-experts\n\nOrchestration, data validation, risk geometry, hypothesis execution, and forensic audit. The infrastructure that makes alpha-squad real - or kills it.\n\n**Agents**:\n\n**`/mft-strategist`** - Obsessive coordinator\n- Orchestrates hypothesis generation and destruction\n- Routes Data Sentinel → Alpha Squad → Factor Geometer → Skeptic\n- Use for: Starting any research, SHIP/KILL/ITERATE decisions\n\n**`/data-sentinel`** - Paranoid gatekeeper\n- MUST be invoked FIRST on any data\n- Trusts nothing, asks before filtering\n- Use for: Any data validation\n\n**`/factor-geometer`** - Risk architect\n- Factor loadings, covariance estimation, alpha-orthogonal decomposition\n- Use for: Geometry check, factor exposure analysis\n\n**`/skeptic`** - Hypothesis executioner\n- Dual causal + statistical validation gauntlet\n- Use for: Full hypothesis validation (Rademacher, walk-forward, placebos)\n\n**`/forensic-auditor`** - Post-mortem investigator\n- Traces assumption failures, runs on schedule and crisis\n- Use for: Pipeline review, incident investigation, periodic audit\n\n**Details**: [mft-research-experts/README.md](mft-research-experts/README.md)\n\n```bash\n/plugin install mft-research-experts@deevs-agent-system\n```\n\n---\n\n### bug-hunters\n\nSystematic bug hunting with spec reconstruction, adversarial validation, and confidence-ranked reports.\n\n**Agents**:\n\n**`/orchestrator`** - Central brain (RED)\n- Reconstructs spec from code, spawns hunters, challenges findings\n- Filters false positives via adversarial validation with dev-experts\n- Use for: Starting any bug hunting session\n\n**`/logic-hunter`** - Spec detective (ORANGE)\n- Language-agnostic logic bugs, spec-vs-implementation gaps\n- Scan mode (hotspots) → Hunt mode (deep trace)\n- Use for: Algorithm correctness, data flow issues, design intent verification\n\n**`/cpp-hunter`** - C++ bug hunter (YELLOW)\n- Memory corruption, UB, concurrency issues\n- Hypothesis-driven, demands proof\n- Use for: Crash debugging, memory bugs, mysterious C++ failures\n\n**`/python-hunter`** - Python bug hunter (YELLOW)\n- Async pitfalls, None propagation, type violations\n- Hypothesis-driven, demands proof\n- Use for: Python-specific bugs, type mismatches, async issues\n\n**Flow**: Spec Reconstruction → Hotspot Scan → Hunter Deployment → Adversarial Challenge → Confidence Scoring → Report\n\n**Details**: [bug-hunters/README.md](bug-hunters/README.md)\n\n```bash\n/plugin install bug-hunters@deevs-agent-system\n```\n\n---\n\n### cost-status\n\nStatus bar showing session cost, monthly cost, total cost, and context window usage.\n\n**Display**: `$0.52/$30.00/$150.00 | 38k/200k (19%)`\n\n- `$session/$month/$total` - Cost tracking at three levels\n- `used/max (%)` - Context window usage\n\n**Setup**: After installation, add to `~/.claude/settings.json`:\n```json\n{\n  \"statusLine\": {\n     \"type\": \"command\",\n     \"command\": \"bash ~/.claude/scripts/show-cost.sh\"\n   },\n}\n```\n\nFind the path with:\n```bash\nfind ~/.claude -name \"show-cost.sh\" -path \"*/cost-status/*\" 2>/dev/null | head -1\n```\n\n**Details**: [cost-status/README.md](cost-status/README.md)\n\n\n```bash\n/plugin install cost-status@deevs-agent-system\n```\n\n### arxiv-search\n\nSearch arXiv preprint repository for academic papers.\n\n**Core**: Query arXiv for research papers across physics, mathematics, computer science, quantitative biology, finance, and statistics.\n\n**Quick start**:\n```bash\n# Basic search (auto-selects Python or bash)\narxiv_search \"transformer attention mechanism\"\n\n# Limit results\narxiv_search \"protein folding\" --max-papers 5\n```\n\n**Use when**: Finding preprints, ML/AI papers, mathematical methods, scientific literature before journal publication.\n\n**Dependencies**: None (bash fallback). For better reliability: `uv pip install arxiv`\n\n**Details**: [arxiv-search/SKILL.md](arxiv-search/SKILL.md)\n\n```bash\n/plugin install arxiv-search@deevs-agent-system\n```\n\n## Agent Color Scheme\n\nUniversal color scheme across all agent plugins:\n\n| Color | Role | Examples |\n|-------|------|----------|\n| ❤️ **RED** | Deciders & Orchestrators | `architect`, `mft-strategist`, `orchestrator`, `devops` |\n| 🧡 **ORANGE** | Hybrid (can lead or challenge) | `logic-hunter` |\n| 💛 **YELLOW** | Checkers & Validators | `reviewer`, `cpp-hunter`, `python-hunter`, `skeptic`, `forensic-auditor` |\n| 💙 **BLUE** | Builders & Implementers | `cpp-dev`, `python-dev`, `rust-dev`, `tester` |\n| 💚 **CYAN** | Researchers & Analysts | `data-sentinel`, `fundamentalist`, `vulture`, `network-architect`, `book-physicist`, `causal-detective`, `factor-geometer` |\n\n## Credits\n\nInspired by:\n- [claude-code-tools](https://github.com/pchalasani/claude-code-tools/) by pchalasani\n- [superpowers](https://github.com/obra/superpowers/) by obra\n- [agent-commands](https://github.com/mitsuhiko/agent-commands) by mitsuhiko\n",
  "bytes": 9734,
  "sha": "4014965e2853140237c03136238c97d3f3a61f7c7ba539775369229602663d35",
  "repo_slug": "deevsdeevs/agent-system",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/skl_deevsdeevs_agent_system_datetime_main_9da8dc9f/readme"
}