AgentSkillsCN

validate-idea

评估单个想法,并决定是接受还是拒绝。此技能由 /validate-ideas 内部使用,或在用户提及特定想法编号时使用。

SKILL.md
--- frontmatter
name: validate-idea
description: Evaluate a single idea and decide to accept or reject it. Used internally by /validate-ideas or when the user mentions a specific idea number.

Validate Idea

Evaluate a pending idea and decide whether to accept or reject it.

Instructions

When validating idea #N:

Step 1 - Fetch the Idea

bash
gh issue view N --repo jmlweb/tooling --json number,title,body,labels

Verify it has the idea:pending label. If not, inform user.

Step 2 - Analyze the Idea

Evaluate against project context:

  1. Read relevant files:

    • README.md, AGENTS.md for project goals
    • Existing packages and their purpose
  2. Check for duplicates:

    • Search existing issues for similar concepts
    • Check if functionality already exists in codebase
  3. Assess feasibility:

    • Technical complexity
    • Dependencies required
    • Integration with existing architecture

Step 3 - Estimate Effort and Impact

Effort levels:

LevelTimeScope
Low<2 hoursSingle file, minimal testing
Medium2-8 hrsMultiple files, moderate testing
High>8 hoursSignificant changes, extensive

Impact levels:

LevelDescription
LowNice to have, minimal user benefit
MediumNoticeable improvement, moderate benefit
HighGame changer, significant benefit

Step 4 - Apply Decision

If ACCEPTED:

bash
gh issue edit N --repo jmlweb/tooling \
  --remove-label "idea:pending" \
  --add-label "idea:accepted,effort:medium,impact:high"

gh issue comment N --repo jmlweb/tooling --body "## Validation: ACCEPTED

**Effort:** [level] | **Impact:** [level]

**Reasoning:** [Why accepted]

*Ready for /feed-backlog*"

If REJECTED:

bash
gh issue edit N --repo jmlweb/tooling \
  --remove-label "idea:pending" \
  --add-label "idea:rejected"

gh issue comment N --repo jmlweb/tooling --body "## Validation: REJECTED

**Reason:** [Why rejected]"

gh issue close N --repo jmlweb/tooling

Decision Guidelines

Accept when: High impact + low effort, aligns with goals, solves pain point

Reject when: Low ROI, out of scope, duplicates, infeasible