spec-explore

Original：🇨🇳 Chinese

Translated

Information collection and exploration before Spec creation. Invoked by the role spec-explorer. Trigger conditions: (1) The spec-explorer role needs to collect background information before Spec creation, (2) Need to retrieve historical experience (exp-search), (3) Need to explore the project codebase, external documents or third-party libraries, (4) TeamLead notifies the spec-explorer to start working.

1installs

Sourcehhu3637kr/skills

Added on2026-03-07

NPX Install

npx skill4agent add hhu3637kr/skills spec-explore

SKILL.md Content (Chinese)

View Translation Comparison →

Spec Explore

Core Principles

Explore First: Complete before spec-writer writes plan.md, providing sufficient background for spec-writer and spec-tester
Conditional exp-reflect: Only invoke exp-reflect when exploring new content; do not invoke when only retrieving existing experience
exploration-report.md is the output: A structured report for reference by spec-writer and spec-tester

exp-reflect Trigger Conditions

Scenario	Trigger exp-reflect?
Invoke exp-search to retrieve existing experience	❌ Do not trigger
Explore the project's existing codebase	✅ Trigger (knowledge memorization)
Explore external codebases (e.g., third-party libraries)	✅ Trigger (knowledge memorization)
Read external technical documents	✅ Trigger (knowledge memorization)

Workflow

Step 1: Receive Task

Obtain from TeamLead's SendMessage:

Current task description
Scope of exploration (project code, external libraries, documents, etc.)
Save path for exploration-report.md

Step 2: Retrieve Historical Experience

bash

/exp-search <keywords>

Retrieve using task keywords, read relevant experience, and record referenceable historical solutions.

Step 3: Explore Project Status

Explore relevant code as needed for the task:

Use
```
Grep
```
to find relevant code patterns
Use
```
Read
```
to read key files
Use
```
Glob
```
to understand directory structure

[!tip] Trigger exp-reflect after exploration After exploring project code, invoke exp-reflect to deposit new findings as knowledge memories.

Step 4: Explore External Resources (if needed)

Explore external codebases (e.g., AgentScope, third-party libraries) or technical documents.

[!tip] Trigger exp-reflect after exploration After exploring external resources, invoke exp-reflect to deposit key findings as knowledge memories.

Step 5: Produce exploration-report.md

Create

exploration-report.md

in the same directory as plan.md:

markdown

# Exploration Report

## Retrieved Historical Experience
[Summary of exp-search results, including experience links]

## Project Status Analysis
[Understanding of relevant code/modules: current implementation, interfaces, data structures]

## External Knowledge
[Key findings from exploring external libraries/documents]

## Recommendations for Spec Creation
[Reference information for spec-writer and spec-tester:
- Recommended implementation directions
- Known boundary cases and risks
- Reusable existing components]

Step 6: Notify spec-writer and spec-tester

python

SendMessage(
    recipient="spec-writer",
    content="exploration-report.md has been completed. Please read it before writing plan.md. Path: {path}"
)
SendMessage(
    recipient="spec-tester",
    content="exploration-report.md has been completed. Please read it and collaborate with spec-writer to write test-plan.md. Path: {path}"
)

Collaboration with Other Roles

TeamLead → spec-explorer starts
spec-explorer → exp-search (retrieval) + code exploration + external resource exploration
spec-explorer → exploration-report.md
spec-explorer → Notify spec-writer + spec-tester

Follow-up Actions

After completing exploration, confirm:

exploration-report.md has been created in the correct path
exp-reflect has been invoked to deposit knowledge after exploring new content
spec-writer and spec-tester have been notified

Common Pitfalls

Mistakenly triggering exp-reflect after invoking exp-search (should not trigger)
exploration-report.md content is too brief, leaving spec-writer without sufficient background information
Not notifying spec-tester, leading to disconnection between test plan and design scheme