Agent Workspace

Pack comparison

Two packs, side-by-side. Merged comparisons, shared shape, and diff highlights in one view.

ACommunityharnessorchestrator-workersv0.1.0Recommended
Turn Execution Pipeline

turn-execution-pipeline

The 9-step loop, 5 context shapers, 3 recovery paths. The 98.4% infrastructure under every turn.

npx attrition-sh pack install turn-execution-pipeline

Token budget

200,000

Pass rate

Avg tokens

Publisher

Agent Workspace

claude-codecursorcodex

3 required outputs, 4 permissions, 3 completion conditions.

out: .transcripts/<session>.jsonl

BCommunityharnessevaluator-optimizerv0.1.0Production-ready
Claude Advisor Pattern v2

advisor-pattern-v2

Sonnet executes, Opus advises. Route intelligence by confidence, pay Opus only when it matters.

npx attrition-sh pack install advisor-pattern-v2

Token budget

12,000

Pass rate

Avg tokens

Publisher

Agent Workspace

claude-codecursorpython-3.11node-20

4 required outputs, 5 permissions, 5 completion conditions.

out: .advisor/<task-id>.json

What both packs have in common

Overlap across canonical pattern, compatibility, tags, and required packs.

claude-codecursor
harness

Head-to-head claims from both packs

Each row is attributed to the pack that authored it. The winner column is normalised to this compare view (A / B / Tie).

SourceAlternativeAxisWinnerNote
Aadvisor-pattern-v2complexityAlternativeAdvisor is a specialized loop shape running inside a pipeline like this one. Advisor is simpler because it ignores the shaper chain; this pack is the whole pipeline.
Aclaude-code-guidemaintainabilityTieClaude Code Guide is the operator's view (AGENTS.md, hooks, skills); this pack is the implementer's view (loop, shapers, recovery). Use together when onboarding a team building a clone.
Aorchestrator-workersaccuracyTieOrchestrator-workers is the canonical pattern family; this pack is the specific instantiation inside Claude Code's harness. The match is substantial but not total — the 5-shaper chain is not part of the generic orchestrator-workers shape.
Bsingle-model-sonnetaccuracyBAdvisor retains ~93% of Opus-only pass-rate vs Sonnet-only's ~85%. The delta comes from the 8-12% of steps that actually need Opus.
Bsingle-model-sonnetcostAlternativeSonnet-only is ~20% cheaper than the advisor variant on the same workload — if the quality gap doesn't matter, skip the advisor.
Bprompt-chainingcomplexityAlternativePrompt chaining is simpler to implement but can't route by difficulty. Advisor adds complexity to capture the cost/quality tradeoff.
Borchestrator-workerslatencyBAdvisor is a flat loop — no orchestrator fan-out. Lower tail latency on sequential tasks. Orchestrator wins on parallelisable ones.

What each pack brings that the other doesn't

Unique coverage and any measurable gap between the two.

Comparisons not in B

advisor-pattern-v2claude-code-guide

Compatibility A-only

codex

Tags A-only

turn-loopcontext-compactionorchestrator-workersclaude-code-internalsdive-into-claude-code

Comparisons not in A

single-model-sonnetsingle-model-sonnetprompt-chaining

Compatibility B-only

python-3.11node-20

Tags B-only

evaluator-optimizeradvisorcost-routingopussonnet

Difference: 188,000 tokens. B (Claude Advisor Pattern v2) is the cheaper declared contract.