[{"data":1,"prerenderedAt":1197},["ShallowReactive",2],{"navigation_docs":3,"blog-2026-05-01-sdd-compared":137},[4,28,62,94,100,133],{"title":5,"path":6,"stem":7,"children":8,"order":11,"icon":27},"Getting Started","/getting-started","1.getting-started/index",[9,12,17,22],{"title":10,"path":6,"stem":7,"order":11},"What is Gangsta Agents?",1,{"title":13,"path":14,"stem":15,"order":16},"Installation","/getting-started/installation","1.getting-started/installation",2,{"title":18,"path":19,"stem":20,"order":21},"Quickstart — Your First Heist","/getting-started/quickstart","1.getting-started/quickstart",3,{"title":23,"path":24,"stem":25,"order":26},"The 1% Rule","/getting-started/the-1-percent-rule","1.getting-started/the-1-percent-rule",4,"i-lucide-rocket",{"title":29,"path":30,"stem":31,"children":32,"order":11,"icon":61},"Core Concepts","/core-concepts","2.core-concepts/index",[33,34,39,43,48,52,56],{"title":29,"path":30,"stem":31,"order":11},{"title":35,"path":36,"stem":37,"order":38},"Intent Routing","/core-concepts/intent-routing","2.core-concepts/intent-routing",5,{"title":40,"path":41,"stem":42,"order":26},"Omerta — The Five Laws","/core-concepts/omerta","2.core-concepts/omerta",{"title":44,"path":45,"stem":46,"order":47},"Spec-Driven Development","/core-concepts/spec-driven-dev","2.core-concepts/spec-driven-dev",7,{"title":49,"path":50,"stem":51,"order":16},"The Gangsta Agents Family Hierarchy","/core-concepts/the-borgata","2.core-concepts/the-borgata",{"title":53,"path":54,"stem":55,"order":21},"The Heist Pipeline","/core-concepts/the-heist","2.core-concepts/the-heist",{"title":57,"path":58,"stem":59,"order":60},"The Ledger","/core-concepts/the-ledger","2.core-concepts/the-ledger",6,"i-lucide-brain",{"title":63,"icon":64,"path":65,"stem":66,"children":67},"Heist Pipeline","i-lucide-workflow","/heist-pipeline","3.heist-pipeline",[68,70,74,78,82,86,90],{"title":53,"path":65,"stem":69,"order":11},"3.heist-pipeline/index",{"title":71,"path":72,"stem":73,"order":16},"Reconnaissance","/heist-pipeline/reconnaissance","3.heist-pipeline/1.reconnaissance",{"title":75,"path":76,"stem":77,"order":21},"The Grilling","/heist-pipeline/the-grilling","3.heist-pipeline/2.the-grilling",{"title":79,"path":80,"stem":81,"order":26},"The Sit-Down","/heist-pipeline/the-sit-down","3.heist-pipeline/3.the-sit-down",{"title":83,"path":84,"stem":85,"order":38},"Resource Development","/heist-pipeline/resource-development","3.heist-pipeline/4.resource-development",{"title":87,"path":88,"stem":89,"order":60},"Execution","/heist-pipeline/the-hit","3.heist-pipeline/5.the-hit",{"title":91,"path":92,"stem":93,"order":47},"The Delivery","/heist-pipeline/laundering","3.heist-pipeline/6.laundering",{"title":95,"path":96,"stem":97,"children":98},"Skills Reference","/skills","4.skills/index",[99],{"title":95,"path":96,"stem":97},{"title":101,"icon":102,"path":103,"stem":104,"children":105},"Advanced","i-lucide-zap","/advanced","5.advanced",[106,109,113,117,121,125,129],{"title":107,"path":103,"stem":108,"order":11},"Advanced Topics","5.advanced/index",{"title":110,"path":111,"stem":112,"order":38},"Artifacts Reference","/advanced/artifacts","5.advanced/artifacts",{"title":114,"path":115,"stem":116,"order":60},"Autonomous Mode","/advanced/autonomous-mode","5.advanced/autonomous-mode",{"title":118,"path":119,"stem":120,"order":26},"Checkpoint Recovery","/advanced/checkpoint-recovery","5.advanced/checkpoint-recovery",{"title":122,"path":123,"stem":124,"order":38},"Contributing","/advanced/contributing","5.advanced/contributing",{"title":126,"path":127,"stem":128,"order":16},"Writing Custom Skills","/advanced/custom-skills","5.advanced/custom-skills",{"title":130,"path":131,"stem":132,"order":21},"Multi-Agent Patterns","/advanced/multi-agent-patterns","5.advanced/multi-agent-patterns",{"title":134,"path":135,"stem":136,"order":60},"Glossary","/glossary","6.glossary",{"id":138,"title":139,"body":140,"description":1187,"extension":1188,"links":1189,"meta":1190,"navigation":1192,"path":1193,"seo":1194,"stem":1195,"__hash__":1196},"docs/blog/2026-05-01-sdd-compared.md","Spec-Driven Development in 2026",{"type":141,"value":142,"toc":1169},"minimark",[143,147,150,153,188,191,197,200,205,218,221,226,242,245,271,274,279,289,294,314,319,333,339,344,346,350,366,369,373,376,382,385,391,398,412,416,436,440,454,459,464,466,470,480,483,487,490,516,523,529,535,541,544,548,571,575,583,588,593,595,599,613,616,619,630,636,641,726,732,737,744,748,751,757,760,764,797,801,804,808,811,855,862,868,872,895,899,910,915,920,922,926,930,933,963,967,971,975,1059,1066,1068,1072,1075,1081,1084,1090,1093,1096,1122,1125,1127,1136,1146,1159],[144,145,146],"p",{},"Ask Claude or Copilot to \"add user authentication\" and you'll get something. Whether that something matches what you actually need depends on twenty silent inferences the agent made: which auth pattern, which session strategy, which security level, which error handling. None of them written down. None of them reviewable. All of them yours to debug later.",[144,148,149],{},"This is the structural failure SDD addresses. But \"spec-first\" is one principle implemented four very different ways. GitHub Spec Kit, OpenSpec, BMad Method, and Gangsta Agents agree that the spec comes before the code — they disagree on almost everything else: how strict the gates are, where the spec lives, whether agents debate, whether memory persists across sessions.",[144,151,152],{},"I'll compare them across five axes that actually matter in practice:",[154,155,156,164,170,176,182],"ul",{},[157,158,159,163],"li",{},[160,161,162],"strong",{},"Spec scope"," — feature-level (one change) or project-level (the whole project)",[157,165,166,169],{},[160,167,168],{},"Adversarial review"," — is there a mechanism that attacks a proposal before code is written",[157,171,172,175],{},[160,173,174],{},"Gate strictness"," — can the agent proceed without explicit human approval",[157,177,178,181],{},[160,179,180],{},"Institutional memory"," — is experience preserved between sessions",[157,183,184,187],{},[160,185,186],{},"Cost of entry"," — dependencies, setup time, learning curve",[144,189,190],{},"Ecosystem metrics (stars, community, extensions) live in a separate table — they're useful, but they're not architecture.",[144,192,193],{},[194,195,196],"em",{},"Disclosure: I built Gangsta Agents. It's included for completeness, not promotion. Judge for yourself.",[198,199],"hr",{},[201,202,204],"h2",{"id":203},"github-spec-kit","GitHub Spec Kit",[144,206,207,210,211],{},[160,208,209],{},"88k stars"," | ",[212,213,217],"a",{"href":214,"rel":215},"https://github.com/github/spec-kit",[216],"nofollow","github.com/github/spec-kit",[144,219,220],{},"Official open-source project from GitHub, launched in 2025, with an active ecosystem of extensions and presets.",[144,222,223],{},[160,224,225],{},"How it works:",[144,227,228,229,233,234,237,238,241],{},"Four phases with clear artifacts: Constitution → Specify → Plan → Tasks → Implement. Each phase is triggered via slash commands (",[230,231,232],"code",{},"/speckit.constitution",", ",[230,235,236],{},"/speckit.specify",", etc.) or agent skills. ",[230,239,240],{},"Constitution.md"," is an immutable document with project principles.",[144,243,244],{},"Each phase is a handoff between the human (driving intent) and the AI (doing the writing). You don't move to the next phase until the current artifact is reviewed and accepted:",[154,246,247,253,259,265],{},[157,248,249,252],{},[160,250,251],{},"Specify"," — the developer writes what needs to be built and why. The AI generates a detailed spec covering user stories, acceptance criteria, workflows, and success metrics. No code, no implementation details.",[157,254,255,258],{},[160,256,257],{},"Plan"," — the developer adds technical constraints (tech stack, architecture, compliance, performance). The AI produces a technical plan that respects those constraints. Alternative plans can be generated for comparison.",[157,260,261,264],{},[160,262,263],{},"Tasks"," — the AI breaks work into small reviewable units. Each task is concrete and independently verifiable (\"Create a user registration endpoint that validates email format\" — not \"build authentication\").",[157,266,267,270],{},[160,268,269],{},"Implement"," — the AI executes each task. Code changes are traceable to specific requirements, not to a giant diff with unclear intent.",[144,272,273],{},"The structure is the value. Each phase has a checkpoint where humans catch misalignment before it becomes rework.",[144,275,276],{},[160,277,278],{},"Real workflow:",[280,281,286],"pre",{"className":282,"code":284,"language":285},[283],"language-text","/speckit.constitution  Create principles focused on code quality and testing standards\n/speckit.specify       Build a Kanban board with drag-and-drop, multiple users, no login\n/speckit.plan          Use .NET Aspire, Postgres, Blazor server\n/speckit.tasks\n/speckit.implement\n","text",[230,287,284],{"__ignoreMap":288},"",[144,290,291],{},[160,292,293],{},"Strengths:",[154,295,296,299,302,305,308,311],{},[157,297,298],{},"50+ community extensions: Jira, Azure DevOps, GitHub Issues, worktrees, security review, V-Model, and more",[157,300,301],{},"Presets for customization without touching the core",[157,303,304],{},"Works with 28 AI agents: Claude Code, Gemini CLI, Copilot, Cursor, OpenCode, Codex, Kiro, Windsurf, and others",[157,306,307],{},"Ralph Loop — autonomous execution cycle for fully agentic implementation",[157,309,310],{},"Brownfield bootstrap extension for existing projects",[157,312,313],{},"Enterprise-ready: Azure DevOps, Jira integrations, compliance requirements",[144,315,316],{},[160,317,318],{},"Weaknesses:",[154,320,321,324,327,330],{},[157,322,323],{},"Heavyweight: requires Python, uv, git",[157,325,326],{},"Phase gates exist but aren't strict — agent can continue without explicit approval",[157,328,329],{},"No adversarial review out of the box (extension only)",[157,331,332],{},"Spec lives at feature level, not project level",[144,334,335,338],{},[160,336,337],{},"Best for:"," If you already have Jira or Azure DevOps and need compatibility with your existing process — Spec Kit brings SDD without changing infrastructure.",[144,340,341],{},[194,342,343],{},"Spec Kit is the most polished, most extensible, and most enterprise-ready of the four — but it pays for that with weight and bypassable gates.",[198,345],{},[201,347,349],{"id":348},"openspec","OpenSpec",[144,351,352,210,355,210,360,365],{},[160,353,354],{},"37.7k stars",[212,356,359],{"href":357,"rel":358},"https://github.com/Fission-AI/OpenSpec",[216],"github.com/Fission-AI/OpenSpec",[212,361,364],{"href":362,"rel":363},"https://openspec.dev",[216],"openspec.dev"," | YC W26",[144,367,368],{},"Built by Tabish (@0xTab), YC W26. Philosophy: fluid, not rigid. Iterative, not waterfall.",[144,370,371],{},[160,372,225],{},[144,374,375],{},"Two key actions:",[280,377,380],{"className":378,"code":379,"language":285},[283],"/opsx:propose \"add dark mode\"\n→ creates openspec/changes/add-dark-mode/\n  ✓ proposal.md  — why and what changes\n  ✓ specs/       — requirements and scenarios\n  ✓ design.md    — technical approach\n  ✓ tasks.md     — implementation checklist\n\n/opsx:apply\n→ executes all tasks\n",[230,381,379],{"__ignoreMap":288},[144,383,384],{},"Each change lives in its own folder. Artifacts can be updated at any time — no strict phase transitions.",[144,386,387,390],{},[160,388,389],{},"Spec deltas"," — unique to OpenSpec. Every change produces a diff against the existing spec, not a new document. Reviewers see exactly what's changing in the requirements (added scenarios, modified behaviors, deleted constraints) before any code is written.",[144,392,393,394,397],{},"Example: adding \"Remember me\" to authentication produces a delta showing the new scenario added to ",[230,395,396],{},"auth-session/spec.md",", not a fresh spec file.",[144,399,400,401,404,405,233,408,411],{},"Specs live in the codebase as permanent documentation. Other frameworks treat the spec as planning material that gets discarded after implementation. OpenSpec keeps specs in ",[230,402,403],{},"openspec/specs/"," organized by capability (",[230,406,407],{},"auth-login/",[230,409,410],{},"checkout-cart/","). When a new developer joins, they read the specs to understand what the code is supposed to do — not what it currently does.",[144,413,414],{},[160,415,293],{},[154,417,418,424,427,430,433],{},[157,419,420,421],{},"Minimal overhead: ",[230,422,423],{},"npm install -g @fission-ai/openspec",[157,425,426],{},"Brownfield-first — built for working with existing code",[157,428,429],{},"Iterative approach: spec evolves with the code",[157,431,432],{},"Active community: Discord, 37.7k stars",[157,434,435],{},"Works with 20+ tools",[144,437,438],{},[160,439,318],{},[154,441,442,445,448,451],{},[157,443,444],{},"Fluid means less structure — no strict gates",[157,446,447],{},"No adversarial review",[157,449,450],{},"No institutional memory between sessions",[157,452,453],{},"Spec quality is entirely on the developer",[144,455,456,458],{},[160,457,337],{}," If you want to try SDD today in 5 minutes on an existing project — this is the fastest start of the four.",[144,460,461],{},[194,462,463],{},"OpenSpec treats specs as living code documentation through deltas — the lightest framework here, but you carry the discipline yourself.",[198,465],{},[201,467,469],{"id":468},"bmad-method","BMad Method",[144,471,472,210,475],{},[160,473,474],{},"44.2k stars",[212,476,479],{"href":477,"rel":478},"https://github.com/bmad-code-org/BMAD-METHOD",[216],"github.com/bmad-code-org/BMAD-METHOD",[144,481,482],{},"Full AI-driven framework for the entire development lifecycle — from idea to shipped code. Version 6, actively maintained.",[144,484,485],{},[160,486,225],{},[144,488,489],{},"Specialized named agents guide the project through each phase. Modular architecture — extendable via BMad Builder. Four phases, each producing artifacts that feed the next:",[154,491,492,498,504,510],{},[157,493,494,497],{},[160,495,496],{},"Analysis"," (optional) — brainstorming, domain/market/technical research, product brief. Validates the idea before committing to planning.",[157,499,500,503],{},[160,501,502],{},"Planning"," — PRD with functional and non-functional requirements + UX spec when relevant.",[157,505,506,509],{},[160,507,508],{},"Solutioning"," — architecture document with ADRs, epics broken into stories, and an Implementation Readiness Check that produces a formal gate: PASS / CONCERNS / FAIL.",[157,511,512,515],{},[160,513,514],{},"Implementation"," — sprint planning, story-by-story development, code review, retrospective.",[144,517,518,519,522],{},"For small well-understood work, Quick Flow (",[230,520,521],{},"bmad-quick-dev",") skips phases 1–3 entirely — clarify intent, plan, implement, review in one unified pass.",[144,524,525,528],{},[230,526,527],{},"project-context.md"," — BMad's equivalent of Constitution. A document that captures technology stack, implementation rules, and project conventions. Generated from architecture or codebase, it guides every workflow.",[144,530,531,534],{},[160,532,533],{},"Adversarial Review"," is the recurring quality mechanism, used across the whole pipeline — code review, implementation readiness checks, spec validation. The reviewer agent must find issues — zero findings triggers a halt and re-analysis. The reviewer assumes problems exist and must surface them. Output is structured by severity:",[280,536,539],{"className":537,"code":538,"language":285},[283],"- HIGH   - login.ts:47 - No rate limiting on failed attempts\n- HIGH   - Session token stored in localStorage (XSS vulnerable)\n- MEDIUM - Password validation happens client-side only\n- LOW    - Magic number 3600 should be SESSION_TIMEOUT_SECONDS\n",[230,540,538],{"__ignoreMap":288},[144,542,543],{},"Unlike Gangsta, there is no Nash Equilibrium as a stopping condition — the debate ends at the orchestrator's discretion. Human filtering is required: because the AI is instructed to find problems, false positives are expected. You decide what is real.",[144,545,546],{},[160,547,293],{},[154,549,550,553,556,559,562,565,568],{},[157,551,552],{},"Covers the full lifecycle: analysis → planning → implementation → verification",[157,554,555],{},"12+ specialized named agents (PM, Architect, Developer, UX, Scrum Master, and more)",[157,557,558],{},"Adversarial Review as a built-in feature",[157,560,561],{},"Supports Agile-compatible workflows — iterative sprints, user stories, backlog",[157,563,564],{},"Large community: Discord, YouTube, active GitHub",[157,566,567],{},"Domain-specific modules: Game Dev Studio, Creative Intelligence Suite, Test Architect",[157,569,570],{},"Works with Claude Code, Cursor, Codex CLI",[144,572,573],{},[160,574,318],{},[154,576,577,580],{},[157,578,579],{},"High entry barrier — many concepts to learn upfront",[157,581,582],{},"Quality depends on installed modules (same applies to Spec Kit with its extensions)",[144,584,585,587],{},[160,586,337],{}," If you're building a new product from scratch and want AI at every stage from idea to deployment — BMad covers the full cycle.",[144,589,590],{},[194,591,592],{},"BMad is the only framework that pairs full-lifecycle coverage with a built-in adversarial reviewer — pay the entry cost only if you'll use the whole pipeline.",[198,594],{},[201,596,598],{"id":597},"gangsta-agents","Gangsta Agents",[144,600,601,210,608],{},[160,602,603],{},[212,604,607],{"href":605,"rel":606},"https://github.com/kucherenko/gangsta",[216],"github.com/kucherenko/gangsta",[212,609,612],{"href":610,"rel":611},"https://gangsta.page",[216],"gangsta.page",[144,614,615],{},"Agentic framework for spec-driven development built on the organizational model of the mafia. Zero dependencies, installs in one step. The newest entry of the four — first stable release v1.1.1 in April 2026 — so the comparison reflects an early-stage project rather than an established ecosystem.",[144,617,618],{},"Gangsta is built on the principle of strict hierarchy and explicit gates — every decision is traceable to the contract, every agent answers only to its own level.",[144,620,621,622,625,626,629],{},"Gangsta supports two modes: ",[160,623,624],{},"manual mode"," where the Don approves every phase gate, and ",[160,627,628],{},"autonomous mode"," where the entire pipeline runs end-to-end without human intervention. Choose based on the cost of failure for the task at hand.",[631,632],"img",{"src":633,"alt":634,"style":635},"/pipeline-flow.svg","The Gangsta Agents Heist Pipeline — 6 phases with Don-approved gates","width:100%;border-radius:4px;margin:1.5rem 0 2rem;",[144,637,638],{},[160,639,640],{},"How it works — 6-phase Heist Pipeline:",[642,643,644,660],"table",{},[645,646,647],"thead",{},[648,649,650,654,657],"tr",{},[651,652,653],"th",{},"Phase",[651,655,656],{},"What happens",[651,658,659],{},"Gate",[661,662,663,674,684,694,704,715],"tbody",{},[648,664,665,668,671],{},[666,667,71],"td",{},[666,669,670],{},"Agents scan the codebase, read the Ledger",[666,672,673],{},"Don approves Dossier",[648,675,676,678,681],{},[666,677,75],{},[666,679,680],{},"Adversarial debate: Proposer vs Devils-Advocate, min 2 rounds",[666,682,683],{},"Don approves consensus",[648,685,686,688,691],{},[666,687,79],{},[666,689,690],{},"Formal contract — no code allowed",[666,692,693],{},"Don signs",[648,695,696,698,701],{},[666,697,83],{},[666,699,700],{},"Decomposition into Work Packages",[666,702,703],{},"Don approves War Plan",[648,705,706,709,712],{},[666,707,708],{},"The Hit",[666,710,711],{},"Parallel execution, TDD mandatory",[666,713,714],{},"Don approves completion",[648,716,717,720,723],{},[666,718,719],{},"Laundering",[666,721,722],{},"Verification, integration, Ledger update",[666,724,725],{},"Don declares Heist complete",[144,727,728,731],{},[160,729,730],{},"Strict hierarchy:"," Don (you) → Consigliere → Underboss → Capo → Soldier. No horizontal communication — only through the chain of command.",[733,734,736],"h3",{"id":735},"the-grilling-unique-feature","The Grilling — unique feature",[144,738,739,740,743],{},"Before any spec is written, three agents debate the solution: Proposer proposes the architecture, Devils-Advocate attacks (vulnerabilities, constitution violations, regression risks), Synthesizer integrates valid attacks and the Don's feedback. Minimum 2 rounds, maximum 7, exit only at ",[160,741,742],{},"Nash Equilibrium"," — when the Devils-Advocate can't raise a single new valid objection.",[733,745,747],{"id":746},"the-sit-down-real-example","The Sit-Down — real example",[144,749,750],{},"Here's a fragment from a real contract in gangsta-money-tracker-example:",[280,752,755],{"className":753,"code":754,"language":285},[283],"## Functional Requirements\n- FR-001: Every transaction stores amount in original currency\n- FR-002: Every transaction stores amount_default in home currency\n- FR-003: Every transaction stores exchange_rate at create time\n- FR-008: Changing default currency recalculates ALL existing transactions\n\n## Architectural Decisions\n- AD-001: Integer storage — no float rounding errors\n- AD-002: Currencies table in DB, not hardcoded enum\n- AD-005: Rate stored per-transaction, not computed at query time\n  (exchange rates change; the rate at transaction time is correct)\n",[230,756,754],{"__ignoreMap":288},[144,758,759],{},"No code. Just WHAT and WHY.",[733,761,763],{"id":762},"governance-through-omerta-5-laws","Governance through Omerta — 5 laws:",[765,766,767,773,779,785,791],"ol",{},[157,768,769,772],{},[160,770,771],{},"Introduction Rule"," — no agent-to-agent communication without hierarchy mediation",[157,774,775,778],{},[160,776,777],{},"Rule of Availability"," — all state saved to files before every phase transition",[157,780,781,784],{},[160,782,783],{},"Rule of Truth"," — every claim cites its source. No citation — stronzate",[157,786,787,790],{},[160,788,789],{},"Rule of Tribute"," — token and resource budgets are enforced",[157,792,793,796],{},[160,794,795],{},"Spec is Law"," — if code contradicts the spec, the spec is revised. Never the reverse.",[733,798,800],{"id":799},"institutional-memory","Institutional Memory",[144,802,803],{},"Gangsta stores history in the filesystem. After each Heist the Ledger is updated: successful patterns become Commandments, failures become Negative Constraints in the Constitution. During Reconnaissance, Associates scan the Ledger and include relevant entries in the next Heist's Dossier.",[733,805,807],{"id":806},"dev-skills-separate-toolkit","Dev Skills — separate toolkit",[144,809,810],{},"Beyond the Heist Pipeline, Gangsta ships 7 software development skills that can be invoked anywhere:",[154,812,813,819,825,831,837,843,849],{},[157,814,815,818],{},[160,816,817],{},"Interrogation"," — systematic debugging, finds the root cause before applying fixes",[157,820,821,824],{},[160,822,823],{},"The Drill"," — TDD enforcement, no production code without a failing test first",[157,826,827,830],{},[160,828,829],{},"Safehouse"," — git worktrees, isolated operational bases for parallel work",[157,832,833,836],{},[160,834,835],{},"The Audit"," — code review with adversarial framing",[157,838,839,842],{},[160,840,841],{},"Receiving Orders"," — processes user feedback with rigor, not blind agreement",[157,844,845,848],{},[160,846,847],{},"The Sweep"," — verification through evidence, not assumption",[157,850,851,854],{},[160,852,853],{},"Exit Strategy"," — clean branch integration and cleanup",[144,856,857,858,861],{},"The ",[160,859,860],{},"1% Rule"," governs invocation: if there's even a 1% chance a skill applies, you invoke it. No rationalizing, no shortcuts. This prevents the \"quick fix\" mentality that compounds into technical debt.",[144,863,864,867],{},[160,865,866],{},"Lightweight Heist"," for minor fixes: the full 6-phase pipeline isn't always required. For minor fixes or known patterns, a lightweight version skips early phases. For new features, architectural changes, or non-trivial bugs — the full pipeline runs.",[144,869,870],{},[160,871,293],{},[154,873,874,877,880,883,886,889],{},[157,875,876],{},"The Grilling — the only adversarial debate with Nash Equilibrium and fixed rounds",[157,878,879],{},"Strict gates in manual mode — every phase requires explicit Don approval",[157,881,882],{},"Institutional Memory via Ledger",[157,884,885],{},"Zero dependencies — pure Markdown and shell scripts",[157,887,888],{},"TDD enforced at the architecture level",[157,890,891,892],{},"Native integration: Claude Code, GitHub Copilot, Gemini CLI, OpenCode, Codex. Cursor support via ",[230,893,894],{},"npx skills add",[144,896,897],{},[160,898,318],{},[154,900,901,904,907],{},[157,902,903],{},"Steep learning curve due to metaphors and hierarchy",[157,905,906],{},"6-phase pipeline with gates is overkill for isolated single-agent tasks",[157,908,909],{},"Young project, community still forming",[144,911,912,914],{},[160,913,337],{}," If the cost of failure is high, the feature touches multiple domains, and you need full traceability of every decision back to the contract — Gangsta gives the strictest process of the four.",[144,916,917],{},[194,918,919],{},"Gangsta Agents trades ceremony for traceability — the only framework with Nash Equilibrium debate and institutional memory, but the youngest of the four.",[198,921],{},[201,923,925],{"id":924},"comparison","Comparison",[733,927,929],{"id":928},"autonomous-execution-where-the-control-boundary-sits","Autonomous execution: where the control boundary sits",[144,931,932],{},"All four frameworks support autonomous execution — but the human control boundary differs:",[154,934,935,945,951,957],{},[157,936,937,940,941,944],{},[160,938,939],{},"GitHub Spec Kit — Ralph Loop:"," agent takes ",[230,942,943],{},"tasks.md"," and executes all tasks without stopping. Gates can be bypassed.",[157,946,947,950],{},[160,948,949],{},"OpenSpec — /opsx:apply:"," executes all tasks sequentially. Minimal gates, maximum autonomy.",[157,952,953,956],{},[160,954,955],{},"BMad Method — Orchestrator:"," orchestrator agent guides execution through phases. Adversarial review built in, but no named loop.",[157,958,959,962],{},[160,960,961],{},"Gangsta Agents — Two modes:"," in manual mode, Soldiers execute Work Packages with strict gates before and after the phase; in autonomous mode, the entire 6-phase pipeline runs end-to-end without human approval.",[733,964,966],{"id":965},"technical-characteristics","Technical characteristics",[631,968],{"src":969,"alt":970,"style":635},"/tech.jpeg","Technical characteristics comparison across four SDD frameworks",[733,972,974],{"id":973},"ecosystem","Ecosystem",[642,976,977,992],{},[645,978,979],{},[648,980,981,983,985,987,989],{},[651,982],{},[651,984,204],{},[651,986,349],{},[651,988,469],{},[651,990,991],{},"Gangsta",[661,993,994,1011,1026,1042],{},[648,995,996,999,1002,1005,1008],{},[666,997,998],{},"GitHub stars",[666,1000,1001],{},"88k",[666,1003,1004],{},"37.7k",[666,1006,1007],{},"44.2k",[666,1009,1010],{},"—",[648,1012,1013,1016,1019,1021,1024],{},[666,1014,1015],{},"Extensions",[666,1017,1018],{},"50+",[666,1020,1010],{},[666,1022,1023],{},"Modules",[666,1025,1010],{},[648,1027,1028,1031,1034,1037,1039],{},[666,1029,1030],{},"Community",[666,1032,1033],{},"Large",[666,1035,1036],{},"Growing",[666,1038,1033],{},[666,1040,1041],{},"Forming",[648,1043,1044,1047,1050,1053,1056],{},[666,1045,1046],{},"AI agents",[666,1048,1049],{},"28",[666,1051,1052],{},"20+",[666,1054,1055],{},"12+",[666,1057,1058],{},"5+",[144,1060,1061,1062,1065],{},"Gangsta's Agile compatibility is low ",[160,1063,1064],{},"by design",": each phase requires explicit human approval before proceeding, which conflicts with sprint-based iteration. This is optimized for traceability, not throughput.",[198,1067],{},[201,1069,1071],{"id":1070},"what-all-four-share","What all four share",[144,1073,1074],{},"Despite different approaches, all four frameworks agree on one thing:",[144,1076,1077,1080],{},[160,1078,1079],{},"The human stays in the loop"," — but not for everything. The agent acts on the rules and only stops where a human is actually needed: conflicting requirements, an architectural choice with real consequences, ambiguity affecting the system boundary.",[144,1082,1083],{},"Uncertainty doesn't disappear into confident-sounding text — it gets logged explicitly.",[144,1085,1086,1089],{},[160,1087,1088],{},"Agent context matters more than agent capability."," When Andrej Karpathy named vibe coding in February 2025, the conversation focused on whether models were good enough. Two years later it's clear: the bottleneck isn't model capability. It's that the model has no contract.",[144,1091,1092],{},"SDD does not guarantee correct decisions and does not replace architectural thinking. It reduces the probability of two specific failure modes: agents drifting off-spec due to missing constraints, and contradictions accumulating silently across sessions.",[144,1094,1095],{},"The four frameworks sit on a spectrum from low ceremony to high traceability:",[154,1097,1098,1104,1110,1116],{},[157,1099,1100,1103],{},[160,1101,1102],{},"OpenSpec:"," minimal structure, maximum iteration speed, no gates",[157,1105,1106,1109],{},[160,1107,1108],{},"Spec Kit:"," structured phases, enterprise integrations, bypassable gates",[157,1111,1112,1115],{},[160,1113,1114],{},"BMad:"," full lifecycle, adversarial review, orchestrator-controlled flow",[157,1117,1118,1121],{},[160,1119,1120],{},"Gangsta Agents:"," strict gates at every phase, Nash Equilibrium debate, institutional memory",[144,1123,1124],{},"Pick based on your failure cost and change velocity. Higher failure cost requires stricter gates. Higher change velocity requires lower ceremony.",[198,1126],{},[144,1128,1129,1132,1133],{},[160,1130,1131],{},"GitHub Spec Kit:"," ",[212,1134,217],{"href":214,"rel":1135},[216],[144,1137,1138,1132,1140,210,1143],{},[160,1139,1102],{},[212,1141,359],{"href":357,"rel":1142},[216],[212,1144,364],{"href":362,"rel":1145},[216],[144,1147,1148,1132,1151,210,1154],{},[160,1149,1150],{},"BMad Method:",[212,1152,479],{"href":477,"rel":1153},[216],[212,1155,1158],{"href":1156,"rel":1157},"https://docs.bmad-method.org",[216],"docs.bmad-method.org",[144,1160,1161,1132,1163,210,1166],{},[160,1162,1120],{},[212,1164,607],{"href":605,"rel":1165},[216],[212,1167,612],{"href":610,"rel":1168},[216],{"title":288,"searchDepth":16,"depth":16,"links":1170},[1171,1172,1173,1174,1181,1186],{"id":203,"depth":16,"text":204},{"id":348,"depth":16,"text":349},{"id":468,"depth":16,"text":469},{"id":597,"depth":16,"text":598,"children":1175},[1176,1177,1178,1179,1180],{"id":735,"depth":21,"text":736},{"id":746,"depth":21,"text":747},{"id":762,"depth":21,"text":763},{"id":799,"depth":21,"text":800},{"id":806,"depth":21,"text":807},{"id":924,"depth":16,"text":925,"children":1182},[1183,1184,1185],{"id":928,"depth":21,"text":929},{"id":965,"depth":21,"text":966},{"id":973,"depth":21,"text":974},{"id":1070,"depth":16,"text":1071},"GitHub Spec Kit, OpenSpec, BMad Method, and Gangsta Agents all implement spec-first development but with different philosophies. Here's how they actually differ — and how to choose.","md",null,{"date":1191},"2026-05-01",false,"/blog/2026-05-01-sdd-compared",{"title":139,"description":1187},"blog/2026-05-01-sdd-compared","9pM1tRI5uOO8K51WGO0MNgc-J789YGqH_c9jGfk9D3Q",1777641834003]