Skip to content

Reference

Quick reference documentation, FAQ, and troubleshooting guides.

Quick Reference

Skills Quick Reference

json
{
  "name": "skill-name",
  "description": "Skill functionality description",
  "prompt": "Detailed instructions...",
  "examples": [
    {"input": "...", "output": "..."}
  ]
}

OpenSpec Quick Reference

CommandPurpose
/opsx:propose "description"Create proposal
/opsx:applyApply implementation
/opsx:archiveArchive change
/opsx:verifyVerify implementation

Harness Quick Reference

javascript
// Define test
const test = {
  name: "Test name",
  input: "Input",
  expected: { contains: ["Expected content"] }
};

// Run test
const result = await runTest(test);

FAQ

Skills FAQ

Q: What's the difference between Skill and Tool?

A: Skill guides AI on how to think, Tool lets AI perform actions. Skill = tactical guidance, Tool = execution action.

Q: Can one Skill do multiple things?

A: It's recommended to keep single responsibility. Complex tasks should be split into multiple Skills and used in combination.

Q: Should Skill examples be as many as possible?

A: No. 2-3 high-quality examples are more effective than 10 low-quality ones.

OpenSpec FAQ

Q: Is OpenSpec suitable for personal projects?

A: Might be overkill for simple features, but more valuable for complex features or team collaboration.

Q: Can specifications be modified?

A: Yes! Specs are living documents and should evolve with implementation.

Q: Will AI definitely implement according to specs?

A: Not necessarily. Use /opsx:verify to verify, and manually review critical parts.

Harness FAQ

Q: What's the difference between Harness and unit tests?

A: Harness tests AI output quality (non-deterministic), unit tests test code logic (deterministic).

Q: What's the purpose of Mock Server?

A: Simulates AI API, no real calls needed, saves cost, more stable tests.


Troubleshooting

Skills Common Issues

Issue 1: AI output format inconsistent

Cause: Prompt format requirements not explicit enough

Solution: Explicitly specify output format (Markdown headings, code blocks, etc.)

Issue 2: Skill not working

Cause: File location or format error

Solution:

  • Check if file is in .skills/ directory
  • Verify JSON format
  • Check if name is unique

OpenSpec Common Issues

Issue 1: /opsx:propose has no response

Solution:

  • Ensure in project root directory
  • Check if initialized (openspec init)

Issue 2: Implementation doesn't match spec

Solution:

  • Use /opsx:verify to verify
  • Update spec to be clearer
  • Provide more examples

Harness Common Issues

Issue 1: Tests always fail

Solution:

  • Check Evaluator configuration
  • Consider using semantic matching instead of exact matching
  • Increase tolerance threshold

基于 MIT 许可发布