AgentSkillsCN

verify

验证命令执行是否符合要求。在确认某项命令或技能已正确完成时使用。

SKILL.md
--- frontmatter
name: verify
description: Verify command execution met requirements. Use when confirming a command or skill completed correctly.
argument-hint: "[--last|--command|--report-only]"
category: workflow

/verify

Usage

bash
/verify                       # Verify last command
/verify --command "/test"     # Verify specific command
/verify --report-only         # Report without recommendations

Description

Evaluates how well a previous command execution met its requirements. Produces an alignment score with actionable recommendations.

Behavior

  1. Extract: Identify requirements from the original command
  2. Analyze: Compare execution against requirements
  3. Score: Calculate alignment percentage
  4. Recommend: Suggest improvements if needed

Expected Output

text
User: /verify

Verifying: /test --create

Analyzing execution...

Verification Results:

Overall Alignment: 85%

Category Breakdown:
  Requirement Completeness: 88%
     - 6/7 explicit requirements met
     - Missing: error boundary handling

  Implementation Quality: 92%
     - Code follows best practices
     - Proper error handling

  Output Accuracy: 75%
     - Format deviation: JSON instead of YAML
     - Missing summary statistics

Key Findings:
  1. Missing error boundary for async operations
  2. Output format should be YAML

Recommendations:
  1. HIGH: Add error boundary handling
  2. MEDIUM: Convert output to YAML format

With fixes applied: projected 94%

Report-Only Mode

text
User: /verify --report-only

[Analysis runs...]

Alignment: 85%
  - 2 issues identified
  - See above for details

(Recommendations omitted in report-only mode)

Scoring Categories

CategoryWeightMeasures
Requirement Completeness30%Explicit + implicit requirements
Implementation Quality25%Code standards, practices
Output Accuracy20%Format, data correctness
Performance15%Execution time, efficiency
Extras/Deviations10%Scope creep, additions

Notes

  • Works with any previously executed command
  • Provides actionable fix commands
  • Typical execution: 1-2 minutes