Menu
Unified skill that guides spec creation through structured, interactive process.
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
npx skills add https://github.com/oimiragieo/agent-studio --skill spec-initThe command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
Agent frontmatter enhancements — disallowedTools, mcpServers scoping, fork_eligible field
Context management — pre-compact persistence, threshold alignment, microcompact/circuit breaker detection
Hook enhancements — updatedInput for bash safety prefixes, suppressOutput for verbose blocks, denial-based routing feedback
Based on SOC occupation classification
Showing SKILL.md
| name | spec-init |
| description | Unified skill that guides spec creation through structured, interactive process. |
| version | 1.0.0 |
| model | sonnet |
| invoked_by | user |
| user_invocable | true |
| tools | ["Read","Write","Edit"] |
| verified | true |
| lastVerifiedAt | "2026-02-22T00:00:00.000Z" |
| source | builtin |
| trust_score | 100 |
| provenance_sha | 1f279769b48bd8e3 |
Unified skill that guides spec creation through structured, interactive process.
Wraps these existing skills:
Question: "What are you building?"
Auto-detect from description:
type: featuretype: bugtype: choretype: refactortype: docsInvoke context-compressor (progressive disclosure mode) with adaptive algorithm:
const { AdaptiveQuestioner } = require('.claude/lib/utils/adaptive-discloser.cjs');
const { ContextAccumulator } = require('.claude/lib/utils/context-accumulator.cjs');
// Determine domain from detected type
const domainMap = {
feature: 'general',
bug: 'debugging',
chore: 'general',
refactor: 'architecture',
docs: 'documentation',
};
const domain = domainMap[detectedType] || 'general';
const aq = new AdaptiveQuestioner(domain);
const ca = new ContextAccumulator();
let history = [];
let questionCount = ;
(questionCount < ) {
context = ca.();
result = aq.(context, history);
readiness = aq.(history, context);
(readiness.) {
;
}
answer = ({ : result. });
ca.(result., answer, { domain, : });
history.({ : result., answer });
questionCount++;
}
summary = ca.();
Key Improvements over v1:
Auto-populate spec from answers:
# SPEC: [Feature Name]
## 1. Overview
**Title**: [From question 1]
**Type**: [Detected type]
**Objective**: [User summary]
**User Story**: As a [user type], I want [capability], so that [benefit]
**Acceptance Criteria**: [From question 5]
## 2. Problem Statement
- **Current State**: [From question 1 answers]
- **Pain Points**: [Extracted from answers]
- **Impact**: [Quantified if possible]
## 3. Proposed Solution
- **Approach**: [From user input]
- **Key Features**: [From answers]
- **Scope**: [What's in/out]
## 4. Implementation Approach
- **Phase 1**: [Design/spike if needed]
- **Phase 2**: [Core implementation]
- **Phase 3**: [Testing]
- **Phase 4**: [Documentation]
## 5. Success Metrics
- **Quantitative**: [From question 3]
- **Qualitative**: [User satisfaction]
- **Timeline**: [From question 4]
## 6. Effort Estimate
- **Design**: 1 day
- **Implementation**: 3 days
- **Testing**: 2 days
- **Documentation**: 1 day
- **Total**: 7 days
## 7. Dependencies
- **Required**: [Extracted from context]
- **Blocking**: [What must complete first]
- **Risk**: [Key risks identified]
## 8. Acceptance Criteria Checklist
- [ ] Feature implemented per spec
- [ ] All tests passing
- [ ] Documentation updated
- [ ] No breaking changes
- [ ] Performance targets met
Validate spec against schema:
After spec approved:
Skill({ skill: "plan-generator", args: { specPath: "..." } })Save spec to:
.claude/context/artifacts/specs/[feature-name]-spec-YYYYMMDD.md
Track metadata:
User: "I want to add dark mode to the UI"
spec-init workflow:
1. Detect: type = "feature"
2. Ask: 5 questions about dark mode
3. User answers in <5 minutes
4. Generate spec
5. Validate against schema
6. Store and offer plan generation
User: "There's a memory leak in the scheduler"
spec-init workflow:
1. Detect: type = "bug"
2. Ask: 5 questions (reproduce steps, impact, etc)
3. Generate bug fix spec
4. Suggest acceptance criteria
5. Ready for planner
.claude/context/artifacts/specs/ with the correct naming convention| Anti-Pattern | Why It Fails | Correct Approach |
|---|---|---|
| Asking 10-12 fixed questions | Over-questioning reduces user engagement | Use progressive disclosure; stop at 5-7 questions when context is sufficient |
| Skipping intent type detection | Questions don't adapt to the task type | Always classify the request as feature/bug/chore/refactor/docs first |
| Generating spec without validation | Incomplete specs reach the planner | Validate all required sections before saving the spec |
| Missing track metadata | Spec cannot be tracked or referenced by downstream agents | Always populate trackId, type, status, and created_at fields |
| Saving to wrong location | Specs are not discoverable by other agents | Always save to .claude/context/artifacts/specs/ with standard naming |
Before starting:
Read .claude/context/memory/learnings.md
After completing:
.claude/context/memory/learnings.md.claude/context/memory/issues.md.claude/context/memory/decisions.mdASSUME INTERRUPTION: If it's not in memory, it didn't happen.