dceoy/speckit-agent-skills

Overview

This skill runs the implementation planning workflow using the Spec Kit plan template to generate concrete design artifacts for a feature. It automates setup, context loading, research, data modeling, API contract generation, and agent-context updates, producing a ready implementation plan and supporting files. The workflow enforces constitution checks and quality gates and reports branch and artifact locations at completion.

How this skill works

The skill invokes repository setup scripts to discover FEATURE_SPEC, IMPL_PLAN, SPECS_DIR, and BRANCH, then loads the feature spec and constitution file. It follows the IMPL_PLAN template through Phase 0 (research) and Phase 1 (design): resolving unknowns, creating research.md, extracting entities to data-model.md, generating API contracts, producing quickstart.md, and updating agent-specific context files. It enforces gates: any unresolved clarification or constitution violation stops the run with an ERROR and a report of required fixes. The command ends after Phase 2 planning and outputs paths to all generated artifacts.

When to use it

• You have a completed feature spec and need a technical implementation plan.
• You want automated research and decision consolidation for unspecified technical choices.
• You need machine-readable API contracts and a data model derived from the spec.
• You need to update agent runtime context files to reflect new technologies.
• You want to enforce repository constitution and quality gates before implementation.

Best practices

• Run from the repo root so setup scripts can resolve absolute paths.
• Provide explicit tech preferences or constraints to reduce NEEDS CLARIFICATION items.
• Resolve high-risk unknowns early by dispatching targeted research agents in Phase 0.
• Keep manual agent-context edits only inside designated markers so updates preserve custom notes.
• Treat ERROR gate outputs as actionable blockers and iterate on spec or constraints before rerunning.

Example use cases

• Turn specs/my-feature/spec.md into specs/my-feature/plan.md, research.md, data-model.md, contracts/, and quickstart.md automatically.
• Resolve ambiguous integrations by generating research tasks and consolidating findings in research.md.
• Produce OpenAPI or GraphQL schemas for backend teams from functional requirements and place them in /contracts/.
• Update agent context files for claude, codex, copilot, or gemini runtimes after adding new libraries or patterns.
• Enforce constitution checks and block progress when security or licensing rules are violated.