agents-standards
Standards for authoring SDD plugin agents — frontmatter, self-containment, skill references, and no-user-interaction rules.
Guided requirements gathering through structured questions to create comprehensive specifications. Produces a complete requirements document via non-blocking conversational interaction.
$ npx -y skills add LiorCohen/sdd --skill spec-solicitation --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/spec-solicitationContext preview
The summary Claude sees to decide when to auto-load this skill.
Guided requirements gathering through structured questions to create comprehensive specifications. Produces a complete requirements document via non-blocking conversational interaction.
name: spec-solicitation description: Guided requirements gathering through structured questions to create comprehensive specifications. Produces a complete requirements document via non-blocking conversational interaction. user-invocable: false
Guide users through structured requirements gathering to create comprehensive specifications. This skill is used for **ALL** spec creation - both interactive and external spec paths.
A new session must be able to resume solicitation with ZERO knowledge of prior conversation. All required information is stored in `solicitation-workflow.yaml`:
**CRITICAL: Never lock the user out of reading your output before they can respond.**
**NEVER:**
**ALWAYS:**
When processing a task from external spec decomposition: 1. Load `context.md` (extracted section from external spec with transformation data) 2. Load discovered components from workflow 3. Pre-populate answers where possible from context 4. Ask clarifying questions for gaps 5. User confirms/refines pre-populated content 6. **DO NOT ask about tech stack** - use discovered components
When processing a purely interactive change: 1. No context available 2. Full solicitation flow from scratch
Even if the external spec says nothing about the technical implementation, the solicitation MUST cover all technical aspects. The solicitation adds the technical dimension that product specs lack. **However, component discovery already identifies WHICH components are needed - solicitation asks about HOW they should work, not WHICH ones to use.**
**CRITICAL: Every question asked and every answer received MUST be recorded in the SPEC.md Requirements Discovery section.**
This includes:
Schema: [`schemas/input.schema.json`](./schemas/input.schema.json)
Accepts change ID, workflow ID, optional context path, and resume flag.
Schema: [`schemas/output.schema.json`](./schemas/output.schema.json)
Returns success status, complete SPEC.md content, save path, and review status.
For detailed guidance, read these on-demand:
After spec generation:
1. Save SPEC.md to drafts location via `workflow_state.save_spec()` 2. Update status to `spec_review` via `workflow_state.update_status()` 3. Move to changes/ via `workflow_state.ready_for_review()` 4. Prompt user to review the spec
During spec review, when user provides feedback:
1. Capture feedback in `review_feedback` array of solicitation-workflow.yaml 2. Update the spec based on feedback 3. Save updated spec 4. Continue until user approves
User: /sdd I want to create a new feature Agent: Starting spec solicitation for: user-auth Step 1: Context & Goal What problem does this solve? User: Users can't log in securely Agent: Who is the primary user? User: End users of the application Agent: What's the expected outcome? User: Secure login with session management [... continues through all steps ...] Agent: Generating SPEC.md from collected requirements... Agent: Spec created at: changes/2026/02/05/a1b2c3-user-auth/01-user-auth/SPEC.md Agent: Please review the spec. When satisfied, tell me: /sdd I want to approve the spec
User: /sdd I want to continue Agent: Resuming workflow a1b2c3 (user-auth) Agent: Current item: API Contracts (user-auth-1) Agent: Loading context... Step 1: Context & Goal Based on the provided context, this appears to be about: - Building interfaces for user authentication - Including login, logout, and session refresh Is that correct? (yes/no/modify) User: yes, but also add password reset Agent: Got it. Adding password reset to the requirements. [... continues with context-aware questions ...]
User: /sdd I want to continue Agent: Resuming spec solicitation for: Auth Contracts (user-auth-1) Previously collected: - Problem: User authentication needs service interfaces - Primary user: Interface consumers - Expected outcome: Auth operations interface - Requirements: 3 collected (login, logout, refresh) Continuing from: Step 5 - Acceptance Criteria For the requirement "Users can login with email and password": What acceptance criteria should we have? (Given/When/Then format) User: Given valid credentials, when user submits login form, then they receive a session token [... continues ...]
This skill uses:
Structure for AI-assisted development AI coding assistants are powerful but chaotic. You prompt, you get code, but then what?
Repo: LiorCohen/sdd
Standards for authoring SDD plugin agents — frontmatter, self-containment, skill references, and no-user-interaction rules.
Standards for authoring SDD plugin commands — frontmatter, user interaction, skill/agent invocation, CLI integration, and output formatting.
Create a commit following repository guidelines with proper versioning and changelog updates.
Two-step self-review at every task lifecycle phase. Step 1 (this skill) runs in-context to gather session signals — files read vs grepped, user pushback, build…
D2 diagramming language reference for architecture diagrams, sequence diagrams, grid layouts, SQL tables, and class diagrams. Produces .d2 files rendered via…
Writes and maintains user-facing documentation for the SDD plugin. Proactively detects when docs are out of sync with plugin capabilities.