AIssistLab

V1 Feature Backlog

Release Polish

1. First-run checklist v2 - implemented
   • Show setup progress across workspace, skills path, index, provider, CLI smoke test, and first chat.
   • Keep actions explicit: choose path, rebuild index, open login, test CLI, send chat.
2. Setup Doctor v2 - implemented
   • Add a V1 readiness score derived from existing deterministic checks.
   • Separate “blocks chat” from “quality warning” from “optional improvement”.
   • Keep sanitized diagnostics as an intentional export action rather than an in-page copy button.
3. Chat readiness polish - implemented
   • Disable send when provider is unavailable unless the UI can render a specific actionable error.
   • Show provider mode, runtime source, index state, and last smoke test in a compact pre-send status row.
   • Add better empty-state questions based on current indexed skills.
4. Browser smoke runbook - implemented through V1 spec evidence and release checklist
   • Document a repeatable local smoke path for Settings, Chat, Skills, Editor, and Export.
   • Add screenshots only after implementation, not in the roadmap.

Skill Lifecycle

1. Skill Template Gallery - implemented
   • Templates: reference skill, workflow skill, command-style skill, subagent-backed skill, learning/rubric skill.
   • Each template includes purpose, fields, starter body, and validation expectations.
2. Skill import preview - implemented
   • Support local folder, archive file, and GitHub URL import candidates.
   • Preview files, detected metadata, duplicate names, supporting-file references, and validation warnings before writing.
   • No install action writes until the user confirms the destination.
3. Skill Quality Doctor v2 - implemented
   • Add checks for official-style frontmatter fields, description clarity, trigger examples, oversized context, missing supporting files, suspicious dynamic commands, and weak instructions.
   • Add quality categories: discoverability, safety, maintainability, portability, and context cost.
4. Safe delete and restore - implemented
   • Require exact skill name confirmation.
   • Move deleted skill to a local trash/backup location before permanent removal.
   • Set index state to stale and show restore guidance.

Claude Code Cockpit

1. Ecosystem inventory - implemented
   • Read project-level .claude/skills, .claude/commands, .claude/agents, .mcp.json, .claude/settings.json, .claude/settings.local.json, and plugin-style folders.
   • Report counts, status, and generic warnings.
   • Do not read user-level credential stores or hidden profile state.
2. Config surface explanations - implemented for V1 cockpit scope
   • Show what each surface is for: skills, commands, agents, MCP, settings, hooks, plugins, memory.
   • Explain what is project-shared versus local-only.
3. Setup Doctor integration - implemented
   • Add group Claude Project.
   • Warn about missing SKILL.md, malformed YAML, duplicate names, and local-only settings that appear shareable.
   • Suggest fixes as text only.
4. Reload guidance - implemented
   • Suggest /reload-skills, /reload-plugins, or session restart based on what changed.
   • Do not run Claude Code commands automatically.

Guided Learning

1. Guided Skill Builder - implemented
   • Ask purpose, user trigger, task type, required inputs, success criteria, and safety boundaries.
   • Generate a draft only after the user supplies enough detail.
   • Let the user edit before save.
2. Rubric feedback - implemented
   • Score drafts against discoverability, specificity, examples, safety, and maintainability.
   • Use deterministic checks first.
   • Optional LLM feedback can be added later only after privacy review.
3. Learning templates - implemented
   • Include templates that teach core skill concepts inline.
   • Use coaching language inspired by Claude Education’s guiding approach.
4. Example prompts - implemented
   • Generate suggested test prompts from skill description and examples.
   • Add “try this in Claude Code” guidance without launching Claude automatically.

Diagnostics And Export

1. V1 readiness report - implemented
   • Include workspace/index/provider/skills/ecosystem summaries.
   • Redact sensitive values and full local paths.
2. Export bundle manifest - implemented
   • Include optional skill quality report, ecosystem inventory, index metadata, and sanitized settings summary.
   • Add manifest version and generated-at timestamp.
3. Privacy audit command - implemented as documented release-gate commands
   • Use the release checklist scans to check API responses and docs for secrets, full home paths, and raw auth/config paths.