Orchestrator
-Owns scope, task graph, boundaries, and integration.
-diff --git a/agents/index.html b/agents/index.html index 3b781e1..ce8da42 100644 --- a/agents/index.html +++ b/agents/index.html @@ -1,88 +1 @@ - - -
- - -Subagent workflow
-- Agents work when roles, files, and evidence are bounded. A worktree gives each worker its - own checkout while the orchestrator protects intent. -
-The tree
-├── agent/ui → components + visual states · ├── agent/tests → acceptance + regressions
- · └── agent/docs → guide + examples · merge after each leaf returns a diff and
- evidence
- Owns scope, task graph, boundaries, and integration.
-Owns one coherent slice and one worktree.
-Re-runs gates and reports remaining gaps.
-Handoff
-Subagent workflow
Agents work when roles, files, and evidence are bounded. A worktree gives each worker its own checkout while the orchestrator protects intent.
The tree
├── agent/ui → components + visual states · ├── agent/tests → acceptance + regressions · └── agent/docs → guide + examples · merge after each leaf returns a diff and evidenceOwns scope, task graph, boundaries, and integration.
Owns one coherent slice and one worktree.
Re-runs gates and reports remaining gaps.
Handoff
The short route
-- Start with the map. Then open the one chapter that matches the decision in front of you: - model, agent, worktree, skill, rule, or proof. -
- Take the full field guide → -Capability and effort are separate knobs.
- Open chapter → -Bound roles, handoffs, and worktrees.
- Open chapter → -Capture repeatable decisions in small packages.
- Open chapter → -Connect guidance to enforcement.
- Open chapter → -Compare a strong prompt with skill-enabled work.
- Open lab → -Browse original packages, references, scripts, and improvements.
- Open desk → -The short route
Start with the map. Then open the one chapter that matches the decision in front of you: model, agent, worktree, skill, rule, or proof.
Take the full field guide →Capability and effort are separate knobs.
Open chapter →Bound roles, handoffs, and worktrees.
Open chapter →Capture repeatable decisions in small packages.
Open chapter →Connect guidance to enforcement.
Open chapter →Compare a strong prompt with skill-enabled work.
Open lab →Browse original packages, references, scripts, and improvements.
Open desk →Model routing
-- A model has a capability ceiling. Effort controls how much room it gets to reason. Route - by uncertainty and verification cost. -
-Lookup, small edits, formatting, and transformations with clear checks.
-Normal implementation where the contract is clear but context matters.
-Planning, architecture, security judgment, and hard failures.
-Two knobs
-strong model + high effort → frame ambiguity · light model + low effort → bounded
- execution · raise one knob at a time → compare evidence
- Sequence
-Model routing
A model has a capability ceiling. Effort controls how much room it gets to reason. Route by uncertainty and verification cost.
Lookup, small edits, formatting, and transformations with clear checks.
Normal implementation where the contract is clear but context matters.
Planning, architecture, security judgment, and hard failures.
Two knobs
strong model + high effort → frame ambiguity · light model + low effort → bounded execution · raise one knob at a time → compare evidenceSequence
Start here
-- This guide turns AI work into a shape: frame the problem, choose the model and agent, - isolate changes, teach repeatable decisions, and verify the result. -
-Capability and effort are separate knobs.
- Open chapter → -Bound roles, handoffs, and worktrees.
- Open chapter → -Capture repeatable decisions.
- Open chapter → -Connect guidance to enforcement.
- Open chapter → -Compare prompts and skill-enabled runs.
- Open lab → -Browse original files and improved drafts.
- Open desk → -Start here
This guide turns AI work into a shape: frame the problem, choose the model and agent, isolate changes, teach repeatable decisions, and verify the result.
Capability and effort are separate knobs.
Open chapter →Bound roles, handoffs, and worktrees.
Open chapter →Capture repeatable decisions.
Open chapter →Connect guidance to enforcement.
Open chapter →Compare prompts and skill-enabled runs.
Open lab →Browse original files and improved drafts.
Open desk →