Skip to content

TODOS.md

Latest commit

 

History

History
130 lines (89 loc) · 4.35 KB

TODOS.md

File metadata and controls

130 lines (89 loc) · 4.35 KB

Roadmap

Last updated: 2026-03-29

v0.3.0 — Open Source Ready

CONTRIBUTING.md

  • What: Contributor guide covering setup, PR conventions, agent creation workflow, and skill authoring.
  • Why: Essential for external contributors to understand how to add agents or improve existing skills.
  • Effort: S
  • Status: Done

SECURITY.md

  • What: Replace GitHub template with real vulnerability reporting process.
  • Why: Current file has placeholder version numbers and no actual policy.
  • Effort: S
  • Status: Done

CHANGELOG

  • What: Retroactive changelog from git history, then maintain going forward.
  • Why: Contributors and users need to understand what changed between versions.
  • Effort: S
  • Status: Done

GitHub templates

  • What: Issue templates (bug report, feature request, new agent proposal), PR template, and funding config.
  • Why: Standardizes contributions and makes it easy for new contributors to participate.
  • Effort: S
  • Status: Done

CI pipeline

  • What: GitHub Actions workflow — lint (Trunk), validate plugin structure, test.
  • Why: PRs need automated quality gates before merge. Currently all checks are manual.
  • Effort: M
  • Status: Todo

Prerequisite checker

  • What: Skill or hook that verifies dependencies (Python version, uv/pip) on install.
  • Why: Prevents wasted time when users hit missing tool errors on first run.
  • Effort: S
  • Status: Todo

Post-install smoke test

  • What: After install, optionally verify agent loading and print a summary.
  • Why: Turns "hope it works" into "confirmed working" on first install.
  • Effort: S
  • Depends on: Prerequisite checker

v0.4.0 — Agent Quality

Skill test harness

  • What: Framework for testing skills with sample inputs and expected behavior patterns.
  • Why: As skill count grows (64+), manual testing doesn't scale. Need automated regression.
  • Effort: M

Agent integration tests

  • What: Tests that verify each agent loads, responds to basic prompts, and uses correct tools.
  • Why: Ensures refactors don't silently break agent behavior.
  • Effort: M

Skill quality scoring

  • What: Lint-like tool that scores skill definitions for completeness (frontmatter, workflow steps, anti-patterns, output format).
  • Why: Maintains consistent quality as community contributes new skills.
  • Effort: M

Agent showcase / demos

  • What: Recorded examples (asciicast or markdown) showing each agent handling a real task.
  • Why: Contributors need to see what "good" looks like. Users need to see what's possible.
  • Effort: L

v0.5.0 — Ecosystem

Plugin marketplace metadata

  • What: Rich metadata (screenshots, categories, compatibility tags) for marketplace listing.
  • Why: Better discovery when Claude Code plugin ecosystem grows.
  • Effort: S

Cross-agent memory

  • What: Shared context mechanism so agents can pass findings to each other without Apex relaying everything.
  • Why: System takeover and multi-agent workflows lose context in handoffs today.
  • Effort: L

Community agent template

  • What: Interactive scaffolding (like create-react-app but for agents) that walks through naming, domain, skills.
  • Why: Lowers the barrier for community contributions beyond the current manual template copy.
  • Effort: M

Agent composition API

  • What: Declarative way to define multi-agent workflows (e.g., "run Warden audit, then Relay hardening, then Vigil instrumentation").
  • Why: Currently only Apex can orchestrate. Users should be able to define custom pipelines.
  • Effort: L

Backlog — Ideas

  • Documentation site (Astro/Starlight) with per-agent pages
  • Agent performance benchmarks (token usage, task completion rates)
  • Specialized sub-agents within teams (e.g., Forge could have Terraform vs Pulumi specialists)
  • Localization of agent prompts for non-English teams
  • VS Code extension for agent selection UI

Done

v0.1.0 (2026-03-16)

  • Initial release with 13 agents and 64 skills
  • Plugin-based architecture
  • Marketplace registration

v0.2.0 (2026-03-28)

  • Removed legacy pip CLI
  • Bundled all agents in root plugin for single install
  • Flattened skills to root for discoverability
  • Removed legacy Cloud Run agent from Forge
  • Refined tagline and added "Why This Exists" section