majiayu000 STANDARD OPERATING PROCEDURE
Purpose
Connect systems through well-defined interfaces, mapping, and resilient middleware with observability and fallback paths.
Trigger Conditions
- Positive: Connect two or more platforms/services; Define data contracts and mappings; Need middleware with retries/backoff and monitoring
- Negative: Simple webhook routing (route to hooks-automation); GitHub-specific automation (route to github-integration); Single-API test harness (route to infrastructure/tooling)
Guardrails
- Structure-first: keep SKILL.md aligned with examples/, tests/, and any resources/references so downstream agents always have scaffolding.
- Adversarial validation is mandatory: cover boundary cases, failure paths, and rollback drills before declaring the SOP complete.
- Prompt hygiene: separate hard vs. soft vs. inferred constraints and confirm inferred constraints before acting.
- Explicit confidence ceilings: format as 'Confidence: X.XX (ceiling: TYPE Y.YY)' and never exceed the ceiling for the claim type.
- MCP traceability: tag sessions WHO=operations-{name}-{session_id}, WHY=skill-execution, and capture evidence links in outputs.
- Avoid anti-patterns: undocumented changes, missing rollback paths, skipped tests, or unbounded automation without approvals.
Required Artifacts
- SKILL.md (this SOP)
- examples/ for integrations
- tests/ for contract validation
- resources/ for templates
Execution Phases
Map systems and constraints
- Identify APIs, auth models, and data domains
- Clarify SLAs, throughput, and ordering needs
- Flag compliance/security requirements
Define contracts and flows
- Design payload schemas and transformations
- Specify retries, backoff, idempotency, and error handling
- Plan observability, alerts, and dashboards
Implement connectors
- Build or configure middleware with least privilege
- Add structured logging and correlation IDs
- Test happy-path and failure scenarios
Validate and operate
- Run integration tests and performance checks
- Document runbooks and fallback paths
- Schedule periodic reviews for drift and SLA health
Output Format
- Integration specification with auth, endpoints, and contracts
- Data mapping and transformation rules
- Test matrix with happy-path and failure cases
- Deployment and runbook notes with monitoring
- Risk and rollback plan for connector changes
Validation Checklist
- Auth and permissions validated across systems
- Idempotency/backoff and retries tested
- Monitoring and alerting configured
- Fallback/disablement procedures documented
- Confidence ceiling stated for integration readiness
Confidence: 0.70 (ceiling: inference 0.70) - Platform integration SOP keeps contracts, retries, and observability explicit