Vibe Coding is Dead.
Start Agent Engineering.
One command. /agt classifies your task, dispatches a team of agents, and routes the right Claude model to each — in your voice.
Watch it orchestrate.
Type one command. agentille classifies, plans, dispatches agents, and ships.
$ /agt "refactor the dashboard sidebar to be collapsible"
→ classifying… feature
→ planning · Opus … 3 steps, 2 parallel
→ dispatching · executor ×2 · Sonnet
→ code review · Opus … ✓ no blockers
→ design review · Opus … ✓ 3 viewports · axe clean
✓ done · PR #42 · feature-team (4 teammates · 12m)
Five specialists. Always the right one.
agentille routes tasks to the agent built for them — not a generic chatbot doing everything at once.
planner
Goal-backward plan with parallel-step markers
executor
Headless implementation, atomic commits, adaptive PR/push
code-reviewer
Bugs, security, quality — read-only, severity-classified
design-reviewer
6-pillar visual review, axe-core, AI-design-tell scan
security-reviewer
Secret leaks, injection, auth bypass — severity-classified
Tokens go where they earn the most.
Each model does exactly the job it's priced and capable for. No Opus on boilerplate. No Haiku on architecture.
Opus
Thinks hard so executors don't have to
- Goal-backward planning
- Architecture decisions
- Code review
- Design review
- Security review
Sonnet
Fast, capable, ships the actual code
- All execution & coding
- File edits & commits
- Adaptive integration (PR / push / local)
Haiku
Dirt-cheap, runs at the edges
- Task classification
- Final summary
- Short-circuit no-ops
Team mode — one agent per pane
Opt in with /agt --team feature-team "...". Each teammate gets its own tmux/iTerm2 pane. Costs more — worth it for complex features or high-stakes reviews.
feature-team
2× executor + code-reviewer + design-reviewer. Ship a feature end-to-end.
review-team
code + design + security reviewer. Full audit on a PR before merge.
incident-team
3× executor, competing hypotheses. Race to a fix under pressure.
The old way vs. the one command
Spoiler: agentille does the chaining for you now.
Do it by hand
- 01
figure out the plan yourself - 02
pick the right model for each step - 03
remember to actually switch models - 04
branch + isolate the work - 05
build it - 06
remember to review it - 07
catch the UI regressions yourself - 08
open the PR by hand - 09
pray nothing conflicts
agentille
/agt "build the feature"
✓ done
(the orchestrator handled the rest)
Install the plugin.
Inside Claude Code. Takes two minutes. Then /agt is ready everywhere.
Requires Claude Code (recent version; 2.1.32+ for team mode).
What you get
Orchestration benefits — not config files.
One-command orchestration
Type /agt "task" and the system classifies, plans, dispatches agents, reviews, and summarizes — you don't wire anything.
Right model per job
Opus for planning and review, Sonnet for execution, Haiku for cheap classification. Tokens go where they earn the most.
Parallel-safe worktrees
Each executor branches off your current branch in its own git worktree. Atomic commits, then integrates adaptively — PR, push, or local branch.
Voice-aware from day one
Your profile.json shapes every prompt. Set brutal honesty once — every planner, executor, and reviewer speaks the same way.
Review built in
Code review on every change. Design review whenever UI is touched. Security reviewer available on demand. No extra step.
Audit it, fork it, ship it.
No black box. The entire orchestration is readable code. See exactly how agents are dispatched, models are routed, and your voice is applied.
By @hasuwini77
github.com/hasuwini77/agentille
