By tk-evans01
Generic-purpose subagents and base skills the rest of the harness depends on.
An unprompted generic subagent. Uses Haiku. Intended for tasks that require less thinking and analysis. Good for summarization, research, and tool calls.
An unprompted generic subagent. Uses Opus. Intended for tasks that require deep reasoning, complex analysis, or nuanced judgment.
An unprompted generic subagent. Uses Sonnet. Intended for tasks that require attention to detail, thinking, and analysis.
Use when analyzing a large corpus of text, code, or data that exceeds a single agent's effective context - orchestrates parallel Worker subagents, Critic review subagents, and a final Summarizer subagent with task tracking and failure recovery
Use to decide which generic-purpose subagent (haiku-general, sonnet-general, opus-general) to dispatch
Claude Code plugin marketplace implementing a strict research-plan-stub-test-implement-eval (RPSTIE) workflow with Tier-2 nearly-pure functional discipline.
Status: M0 — skeleton. Plugin manifests, structure, conventions in place. No agents/skills/commands implemented yet.
Rough Idea
|
v
/start-design-plan ---> Design Document (committed)
| (clear context)
v
/start-implementation-plan ---> Phase files + worktree
| (clear context)
v
/execute-implementation-plan ---> per task: stub-author -> test-author -> body-implementor
per phase: code-reviewer -> bug-fixer (3-strike cap)
end: librarian -> code-reviewer -> test-analyst
|
v
/finish-branch ---> merge / PR / discard
Verification: /verify runs the project's configured linters/typecheck/tests. Linter configs are project-owned (set up once per project, not bundled by the harness). No per-language verify skills.
Main thread runs at top-line model (Opus). Subagents default to Sonnet; reviewer escalates to Opus for subtle catches.
| Plugin | Purpose |
|---|---|
tk-foundation | Generic subagents + base skills |
tk-research | Codebase + internet research subagents |
tk-house-style | Tier-2 nearly-pure FP discipline + per-language quick cards |
tk-verify | Thin /verify runner + edit-validation hook |
tk-rpie | RPSTIE workflow commands and subagents |
tk-hooks | Security hardening, CLAUDE.md reminder, skill reinforcement |
| Stage | Agent | Model |
|---|---|---|
| Orchestrate | main thread | opus |
| Research | codebase-investigator, internet-researcher | sonnet |
| Clarify | clarifying-questioner | sonnet |
| Stubs | stub-author | sonnet |
| Tests | test-author | sonnet |
| Bodies | body-implementor | sonnet |
| Review (per phase) | code-reviewer | opus |
| Fix | bug-fixer | sonnet |
| Final tests | test-analyst | sonnet |
| Docs | librarian | sonnet |
Default for projects using this harness. Project may override via .tk-harness/style.md.
See plugins/tk-house-style/skills/nearly-pure-functional/SKILL.md (M3).
/plugin marketplace add https://github.com/TK-Evans01/tk-harness.git
/plugin install tk-rpie@tk-harness
/plugin install tk-house-style@tk-harness
# ... etc
tk-rpie core — agents + key skills + commandstk-house-style — nearly-pure-functional skill + per-lang cardstk-verify — thin /verify command + edit-validation hooktk-hooks — port ed3d hook trioNote: triage gate deferred. Every task currently routes through full RPSTIE.
RPSTIE vs tk-rpie. "RPSTIE" (Research-Plan-Stub-Test-Implement-Eval) is the workflow acronym, capturing all six stages including the explicit Stub and Test split. "tk-rpie" is the plugin name; we kept the shorter form for namespace ergonomics in skill / agent references (
tk-rpie:writing-stubs-and-docs, etc.). The plugin name is not changing.
Inspired by and partially derived from ed3dai/ed3d-plugins by Ed Ropple, which itself derives from obra/superpowers by Jesse Vincent.
Files derived from those projects retain their original MIT license (see per-plugin LICENSE files).
CC-BY-SA-4.0 for original content. MIT for files derived from obra/superpowers (marked per-file or per-plugin).
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Research subagents (codebase, internet, combined) and supporting skills. Used by tk-rpie design and verification phases.
Tier-2 nearly-pure functional house style; per-language quick cards (TS, Python, Rust); testing discipline; FP-primitives glossary.
Cross-cutting hooks: security hardening (PreToolUse/PostToolUse secret leak detection), CLAUDE.md reminder, skill reinforcement.
Research-Plan-Implement-Eval workflow. Stubs/tests/bodies as separate subagents. Worktree-first, per-phase review with capped fix loop.
Verification gate: typecheck, lint, architectural lint, purity checks. Includes /verify command and edit-validation hook.
npx claudepluginhub tk-evans01/tk-harness --plugin tk-foundationUltra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Curate auto-memory, promote learnings to CLAUDE.md and rules, extract proven patterns into reusable skills.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Memory compression system for Claude Code - persist context across sessions