Agentic AI Atlasby a5c.ai
OverviewWikiGraphFor AgentsEdgesSearchWorkspace
/
GitHubDocsDiscord
iiRecord
Agentic AI Atlas · obra/superpowers
page:docs-reference-repos-obra-superpowers-researcha5c.ai
Search record views/
Record · tabs

Available views

II.Record viewspp. 1 - 1
overviewarticlejsongraph
II.
Page JSON

page:docs-reference-repos-obra-superpowers-research

Structured · live

obra/superpowers json

Inspect the normalized record payload exactly as the atlas UI reads it.

File · wiki/docs/reference-repos/obra/superpowers/research.mdCluster · wiki
Record JSON
{
  "id": "page:docs-reference-repos-obra-superpowers-research",
  "_kind": "Page",
  "_file": "wiki/docs/reference-repos/obra/superpowers/research.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "sourcePath": "docs/reference-repos/obra/superpowers/research.md",
    "sourceKind": "repo-docs",
    "title": "obra/superpowers",
    "displayName": "obra/superpowers",
    "slug": "docs/reference-repos/obra/superpowers/research",
    "articlePath": "wiki/docs/reference-repos/obra/superpowers/research.md",
    "article": "\n# obra/superpowers\n\n## Metadata\n- **Stars:** 148,122\n- **License:** MIT\n- **Last pushed:** 2026-04-10 (very actively maintained)\n- **Topics:** (none)\n- **Fork:** No\n- **Author:** Jesse Vincent (obra)\n\n## Archetype: methodology-repo\n\nThe dominant open-source software development methodology for coding agents. A complete Claude Code plugin (v5.0.7) with 14 composable skills covering the full development lifecycle: brainstorming, planning, TDD, debugging, code review, subagent orchestration, and branch management.\n\n## Structure\n- `.claude-plugin/plugin.json` + `marketplace.json` -- official Claude plugin\n- `.codex/`, `.cursor-plugin/`, `.opencode/` -- multi-harness support\n- `CLAUDE.md`, `AGENTS.md`, `GEMINI.md` -- harness-specific instructions\n- `skills/` with 14 SKILL.md-based skills:\n  - brainstorming\n  - writing-plans (+ plan-document-reviewer-prompt.md)\n  - executing-plans\n  - subagent-driven-development (+ implementer-prompt.md, spec-reviewer-prompt.md, code-quality-reviewer-prompt.md)\n  - dispatching-parallel-agents\n  - test-driven-development (+ testing-anti-patterns.md)\n  - systematic-debugging (+ defense-in-depth.md, root-cause-tracing.md, find-polluter.sh, condition-based-waiting.md)\n  - requesting-code-review\n  - receiving-code-review\n  - finishing-a-development-branch\n  - using-git-worktrees\n  - using-superpowers (meta-skill)\n  - verification-before-completion\n  - writing-skills\n- `agents/`, `commands/`, `hooks/` -- plugin infrastructure\n- `gemini-extension.json` -- Gemini CLI extension\n\n## IMPORTANT CONTEXT\nSuperpowers is already installed as a plugin in the babysitter project (skills are visible in the skill list as `superpowers:*`). This repo is the upstream source. The investigation here is about identifying what methodology patterns could be assimilated as babysitter-native processes, beyond what's already available via the plugin.\n\n## Extractable Value\n\n### Processes\n\n**HIGH VALUE -- Full dev methodologies (methodologies/):**\n\n- **TDD quality convergence** (methodologies/): The test-driven-development skill is a rigorous Red-Green-Refactor methodology with \"Iron Law\" enforcement (no production code without failing test first). Includes anti-patterns reference. This maps to a babysitter process with breakpoints at each phase transition (RED->GREEN->REFACTOR) and quality gates.\n\n- **Systematic debugging** (methodologies/): Four-phase debugging methodology (Root Cause Investigation -> Hypothesis Formation -> Targeted Fix -> Verification). Includes supplementary materials on defense-in-depth, root-cause tracing, and condition-based waiting. Maps to a babysitter process with mandatory phase completion gates.\n\n- **Plan-execute-review pipeline** (methodologies/): The writing-plans + executing-plans + verification-before-completion chain describes a complete plan-driven development methodology. Writing-plans includes a plan-document-reviewer-prompt. This is a three-phase babysitter process with breakpoints between planning, execution, and verification.\n\n**MEDIUM VALUE -- Orchestration patterns (specializations/shared/):**\n\n- **Subagent orchestration pattern** (specializations/shared/): The subagent-driven-development skill defines a dispatch-implement-review cycle with two-stage review (spec compliance then code quality). Each stage has dedicated prompts (implementer-prompt.md, spec-reviewer-prompt.md, code-quality-reviewer-prompt.md). This maps to a babysitter parallel.map() process with nested review breakpoints.\n\n- **Parallel agent dispatch** (specializations/shared/): The dispatching-parallel-agents skill describes patterns for identifying independent tasks and farming them out. Maps to babysitter's parallel.all() effect pattern.\n\n**LOWER VALUE -- Workflow patterns:**\n\n- **Code review cycle** (specializations/shared/): requesting-code-review + receiving-code-review describe a bidirectional review process. Could become a babysitter process with breakpoint-based review gates.\n\n- **Branch finalization** (specializations/shared/): finishing-a-development-branch describes the decision tree for squash vs merge vs rebase and cleanup. Useful as a lightweight babysitter process.\n\n### Plugin Ideas\n- **Superpowers methodology bridge plugin:** Since superpowers is already installed as a plugin, the value is in converting its skill-based methodology into deterministic babysitter processes that can be replayed, journal-tracked, and state-cached. A marketplace plugin could provide `babysitter harness:call --process superpowers-tdd` etc.\n\n## Library Mapping\n\n| Extractable Process | Library Status | Action | Existing Path | Target Placement |\n|-------------------|----------------|--------|---------------|------------------|\n| TDD Quality Convergence | UPGRADE | Enhance with Iron Law enforcement and anti-patterns | library/methodologies/superpowers/superpowers-workflow.js | methodologies/superpowers/ (enhancement) |\n| Systematic Debugging | UPGRADE | Enhance with 4-phase methodology and root-cause tracing | library/methodologies/superpowers/superpowers-workflow.js | methodologies/superpowers/ (enhancement) |\n| Plan-Execute-Review Pipeline | UPGRADE | Enhance with dedicated reviewer prompts and phase gates | library/methodologies/superpowers/superpowers-workflow.js | methodologies/superpowers/ (enhancement) |\n| Subagent Orchestration Pattern | UPGRADE | Enhance with dispatch-implement-review cycle | library/methodologies/superpowers/using-superpowers.js | methodologies/superpowers/ (enhancement) |\n| Parallel Agent Dispatch | UPGRADE | Enhance with independent task identification patterns | library/methodologies/superpowers/superpowers-workflow.js | methodologies/superpowers/ (enhancement) |\n| Code Review Cycle | VARIANT | Could be extracted as shared pattern | - | specializations/shared/bidirectional-review.js |\n| Branch Finalization | VARIANT | Could be extracted as shared workflow pattern | - | specializations/shared/branch-finalization.js |\n\n## Plugin Marketplace Mapping\n\n| Plugin Idea | Marketplace Status | Action | Existing Plugin | Target Placement |\n|-------------|-------------------|--------|-----------------|------------------|\n| Superpowers Methodology Bridge | NEW | Convert superpowers skills to deterministic processes | - | plugins/a5c/marketplace/plugins/superpowers-bridge/ |\n\n### SKIP\n- using-superpowers meta-skill (skill-management)\n- writing-skills (skill-management)  \n- using-git-worktrees (utility, not a process methodology)\n- brainstorming (already lightweight enough as a skill, doesn't benefit from process orchestration)\n- The plugin infrastructure (hooks, commands, agents) -- babysitter has its own\n",
    "documents": []
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "page:docs-reference-repos",
      "to": "page:docs-reference-repos-obra-superpowers-research",
      "kind": "contains_page"
    }
  ]
}

Shortcuts

Back to overview
Open graph tab