Spec

SkillFiles & storage

Drives interactive requirement discovery to produce spec files. Use when the user says "spec this", "write specs", "create specs", or "run the spec skill".

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Spec skill

What this skill tells your AI

The instructions your AI receives, as published by codagent-ai/agent-skills in skills/spec/SKILL.md and read by ahel’s review.

Turn the proposal's capabilities into testable behavioral requirements through collaborative discovery. Specifications define observable behavior, not architecture or implementation.

Do not write a capability's spec until you have presented its proposed requirements and scenarios and the user has approved them.

Process

  1. Read the proposal and related existing specifications. Use the proposal's capability names to determine the new or modified spec files. For modified capabilities, locate the current requirement blocks before drafting the delta.
  2. Work through capabilities one at a time. Use codagent:ask-questions to resolve material behavior, boundaries, errors, and edge cases. Ask only questions that affect observable behavior or scope, recommend a default when useful, and do not use a generic approval question as discovery.
  3. Present the complete proposed requirements and scenarios for that capability, including consequential assumptions or defaults inferred from context.
  4. After approval, write the spec file using the format below. Continue until each proposal capability has a corresponding spec.

If behavior depends on an unresolved architectural choice, specify as much as is currently knowable and add <!-- deferred-to-design: <reason> --> to the affected scenario. The design phase will complete or revise it. Do not ask architecture or implementation questions during specification.

Report the relative paths created. Do not invoke another lifecycle skill.

OpenSpec format

  • Use one file per capability: specs/<capability>/spec.md.
  • Write normative requirements with SHALL or MUST.
  • Use ### Requirement: <name> and at least one #### Scenario: <name> per requirement.
  • Scenarios describe observable WHEN/THEN behavior. Exactly four hashes on scenario headings are required by the parser.
  • Avoid scenarios about file contents, configuration structure, or skill text unless those are the actual public contract.

Use delta sections as applicable:

  • ## ADDED Requirements for new behavior.
  • ## MODIFIED Requirements for changed behavior. Copy the entire existing requirement block and all scenarios before editing it.
  • ## REMOVED Requirements with Reason and Migration.
  • ## RENAMED Requirements with FROM and TO.
## ADDED Requirements

### Requirement: <name>
<normative behavior>

#### Scenario: <name>
- **WHEN** <condition>
- **THEN** <observable result>

Signals

GitHub stars
30
Forks
1
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
spec-codagent-ai
Source
github.com/codagent-ai/agent-skills