saga-orchestration
Implement saga patterns for distributed transactions and cross-aggregate workflows. Use when coordinating multi-step business processes, handling compensating transactions, or managing long-running workflows.
Overall
score
73%
Does it follow best practices?
If you maintain this skill, you can automatically optimize it using the tessl CLI to improve its score:
npx tessl skill review --optimize ./path/to/skillEvaluation — 100%
↑ 1.43xAgent success when using this skill
Validation for skill structure
Evaluation results
100%
57%Patient Onboarding Saga
Orchestration saga implementation
SagaState enum
63%
100%
SagaStep fields
63%
100%
Saga dataclass fields
25%
100%
UUID for saga_id
0%
100%
ABC base orchestrator
13%
100%
Concrete saga extends base
83%
100%
Compensation in reverse order
100%
100%
Skip non-completed steps
88%
100%
COMPENSATING state set
100%
100%
FAILED after all compensated
0%
100%
Event payload saga_id and step_name
0%
100%
Compensation includes original_result
0%
100%
Completion and failure events
0%
100%
Without context: $0.3569 · 1m 37s · 12 turns · 12 in / 5,981 out tokens
With context: $0.7463 · 2m 41s · 26 turns · 25 in / 9,913 out tokens
100%
7%Returns Processing Workflow
Choreography saga with event subscriptions
SagaContext dataclass
67%
100%
Event subscription wiring
100%
100%
Forward path via event publish
100%
100%
Compensation event subscriptions
100%
100%
Compensation publishes rollback events
100%
100%
saga_id propagated
70%
100%
No direct service coupling
100%
100%
In-memory event bus
100%
100%
Event flow documented
100%
100%
Without context: $0.4886 · 2m 16s · 12 turns · 12 in / 10,056 out tokens
With context: $1.0588 · 4m 7s · 23 turns · 23 in / 18,394 out tokens
100%
25%Fund Transfer Saga with Bounded Wait Times
Saga step timeouts and failure handling
Timeout orchestrator class
100%
100%
timeout_at set to 5 minutes
0%
100%
Scheduler.schedule called
100%
100%
Unique timeout job key
100%
100%
Timeout check inspects executing status
83%
100%
handle_step_failed called on timeout
42%
100%
COMPENSATING state on failure
100%
100%
Reverse-order compensation
100%
100%
FAILED after all compensated
60%
100%
Timeout design documented
100%
100%
Without context: $0.5119 · 2m 13s · 19 turns · 19 in / 8,006 out tokens
With context: $1.2174 · 4m 32s · 36 turns · 274 in / 18,079 out tokens
- Evaluated
- Agent
- Claude Code
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.