diff --git a/claude/idea-evaluator/README.md b/claude/idea-evaluator/README.md new file mode 100644 index 0000000..a68618b --- /dev/null +++ b/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//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 + diff --git a/claude/idea-evaluator/SKILL.md b/claude/idea-evaluator/SKILL.md new file mode 100644 index 0000000..b1e82d3 --- /dev/null +++ b/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 + diff --git a/codex/idea-evaluator/README.md b/codex/idea-evaluator/README.md new file mode 100644 index 0000000..61e9d9c --- /dev/null +++ b/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 + diff --git a/codex/idea-evaluator/SKILL.md b/codex/idea-evaluator/SKILL.md new file mode 100644 index 0000000..b1e82d3 --- /dev/null +++ b/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 + diff --git a/openclaw/idea-evaluator/README.md b/openclaw/idea-evaluator/README.md new file mode 100644 index 0000000..91bc3de --- /dev/null +++ b/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 + diff --git a/openclaw/idea-evaluator/SKILL.md b/openclaw/idea-evaluator/SKILL.md new file mode 100644 index 0000000..b1e82d3 --- /dev/null +++ b/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 +