Added Harness skills

claude/idea-evaluator/README.md

@@ -0,0 +1,15 @@
+# Claude Code Package
+
+This folder contains the Claude Code version of `idea-evaluator`.
+
+## Packaging notes
+
+- Claude Code discovers skills from `.claude/skills/<name>/SKILL.md`.
+- The folder name should stay lowercase and hyphenated.
+- `SKILL.md` should stay concise so the loaded instructions remain easy to apply.
+
+## Source
+
+- `Intent.md` in the repo root
+- Claude Code Agent Skills documentation
+

claude/idea-evaluator/SKILL.md

@@ -0,0 +1,90 @@
+---
+name: idea-evaluator
+description: Evaluate startup ideas, product concepts, MVPs, and business ideas with staged multi-agent debate, market reasoning, competitive analysis, and a final score and verdict. Use when asked whether an idea is worth pursuing, whether there is a real market, what the strongest reasons for and against are, or whether to plan an MVP.
+---
+
+# Idea Evaluator
+
+## Purpose
+
+Use this skill to judge whether a user idea is worth building.
+
+Treat the task like a structured decision engine:
+
+- gather opposing viewpoints,
+- pressure-test the idea,
+- assess market potential,
+- and give a clear recommendation.
+
+## Operating Model
+
+Run the evaluation in 3 stages.
+
+### Stage 1: Broad Scan
+
+Launch 5 subagents.
+
+Assign them distinct roles:
+
+- 2 against the idea
+- 2 for the idea
+- 1 neutral agent that gives 1 strong pro and 1 strong con
+
+Make the roles meaningfully different. Do not let them collapse into the same answer.
+
+Each subagent should focus on:
+
+- market need
+- customer pain
+- competition
+- timing
+- feasibility
+- evidence where available
+
+### Stage 2: Deep Debate
+
+Launch 2 additional subagents.
+
+Assign them these roles:
+
+- 1 fully pro
+- 1 fully against
+
+Ask each one to make the strongest possible case for its side.
+
+They should sharpen the debate, not repeat Stage 1.
+
+### Stage 3: Final Judgment
+
+Read all subagent output and make the final call.
+
+Your final response must:
+
+- give the idea a score from 1 to 10,
+- summarize the strongest positive arguments,
+- summarize the strongest negative arguments,
+- state a clear verdict,
+- and ask whether the user wants to plan an MVP next.
+
+## Decision Standards
+
+Use evidence and market logic over vague enthusiasm.
+
+If the market looks weak, crowded, or unproven, say so plainly.
+
+If the idea looks strong, explain why it survives criticism.
+
+Avoid false certainty.
+
+## Final Output Format
+
+Keep the response concise and decision-oriented.
+
+Include:
+
+- score out of 10
+- short verdict
+- positives
+- negatives
+- MVP question
+

codex/idea-evaluator/README.md

@@ -0,0 +1,15 @@
+# Codex Package
+
+This folder contains the Codex version of `idea-evaluator`.
+
+## Packaging notes
+
+- Codex skills are folder-based and use `SKILL.md` as the trigger and instruction file.
+- `name` and `description` in YAML frontmatter are the important discovery fields.
+- Keeping the body procedural makes the skill easier for another Codex instance to reuse.
+
+## Source
+
+- `Intent.md` in the repo root
+- Codex skill format guidance and the OpenAI skill standard
+

codex/idea-evaluator/SKILL.md

@@ -0,0 +1,90 @@
+---
+name: idea-evaluator
+description: Evaluate startup ideas, product concepts, MVPs, and business ideas with staged multi-agent debate, market reasoning, competitive analysis, and a final score and verdict. Use when asked whether an idea is worth pursuing, whether there is a real market, what the strongest reasons for and against are, or whether to plan an MVP.
+---
+
+# Idea Evaluator
+
+## Purpose
+
+Use this skill to judge whether a user idea is worth building.
+
+Treat the task like a structured decision engine:
+
+- gather opposing viewpoints,
+- pressure-test the idea,
+- assess market potential,
+- and give a clear recommendation.
+
+## Operating Model
+
+Run the evaluation in 3 stages.
+
+### Stage 1: Broad Scan
+
+Launch 5 subagents.
+
+Assign them distinct roles:
+
+- 2 against the idea
+- 2 for the idea
+- 1 neutral agent that gives 1 strong pro and 1 strong con
+
+Make the roles meaningfully different. Do not let them collapse into the same answer.
+
+Each subagent should focus on:
+
+- market need
+- customer pain
+- competition
+- timing
+- feasibility
+- evidence where available
+
+### Stage 2: Deep Debate
+
+Launch 2 additional subagents.
+
+Assign them these roles:
+
+- 1 fully pro
+- 1 fully against
+
+Ask each one to make the strongest possible case for its side.
+
+They should sharpen the debate, not repeat Stage 1.
+
+### Stage 3: Final Judgment
+
+Read all subagent output and make the final call.
+
+Your final response must:
+
+- give the idea a score from 1 to 10,
+- summarize the strongest positive arguments,
+- summarize the strongest negative arguments,
+- state a clear verdict,
+- and ask whether the user wants to plan an MVP next.
+
+## Decision Standards
+
+Use evidence and market logic over vague enthusiasm.
+
+If the market looks weak, crowded, or unproven, say so plainly.
+
+If the idea looks strong, explain why it survives criticism.
+
+Avoid false certainty.
+
+## Final Output Format
+
+Keep the response concise and decision-oriented.
+
+Include:
+
+- score out of 10
+- short verdict
+- positives
+- negatives
+- MVP question
+

openclaw/idea-evaluator/README.md

@@ -0,0 +1,15 @@
+# OpenClaw Package
+
+This folder contains the OpenClaw version of `idea-evaluator`.
+
+## Packaging notes
+
+- OpenClaw loads skills from `SKILL.md` files with YAML frontmatter.
+- The skill name lives in frontmatter and should stay lowercase hyphen-case.
+- The body stays concise because it is the operating playbook once the skill triggers.
+
+## Source
+
+- `Intent.md` in the repo root
+- OpenClaw skill loading and frontmatter rules
+

openclaw/idea-evaluator/SKILL.md

@@ -0,0 +1,90 @@
+---
+name: idea-evaluator
+description: Evaluate startup ideas, product concepts, MVPs, and business ideas with staged multi-agent debate, market reasoning, competitive analysis, and a final score and verdict. Use when asked whether an idea is worth pursuing, whether there is a real market, what the strongest reasons for and against are, or whether to plan an MVP.
+---
+
+# Idea Evaluator
+
+## Purpose
+
+Use this skill to judge whether a user idea is worth building.
+
+Treat the task like a structured decision engine:
+
+- gather opposing viewpoints,
+- pressure-test the idea,
+- assess market potential,
+- and give a clear recommendation.
+
+## Operating Model
+
+Run the evaluation in 3 stages.
+
+### Stage 1: Broad Scan
+
+Launch 5 subagents.
+
+Assign them distinct roles:
+
+- 2 against the idea
+- 2 for the idea
+- 1 neutral agent that gives 1 strong pro and 1 strong con
+
+Make the roles meaningfully different. Do not let them collapse into the same answer.
+
+Each subagent should focus on:
+
+- market need
+- customer pain
+- competition
+- timing
+- feasibility
+- evidence where available
+
+### Stage 2: Deep Debate
+
+Launch 2 additional subagents.
+
+Assign them these roles:
+
+- 1 fully pro
+- 1 fully against
+
+Ask each one to make the strongest possible case for its side.
+
+They should sharpen the debate, not repeat Stage 1.
+
+### Stage 3: Final Judgment
+
+Read all subagent output and make the final call.
+
+Your final response must:
+
+- give the idea a score from 1 to 10,
+- summarize the strongest positive arguments,
+- summarize the strongest negative arguments,
+- state a clear verdict,
+- and ask whether the user wants to plan an MVP next.
+
+## Decision Standards
+
+Use evidence and market logic over vague enthusiasm.
+
+If the market looks weak, crowded, or unproven, say so plainly.
+
+If the idea looks strong, explain why it survives criticism.
+
+Avoid false certainty.
+
+## Final Output Format
+
+Keep the response concise and decision-oriented.
+
+Include:
+
+- score out of 10
+- short verdict
+- positives
+- negatives
+- MVP question
+