cloudflare（Cloudflare 平台开发）

Cloudflare Infrastructure Operations

Manage Cloudflare services: Workers, KV, R2, D1, Hyperdrive, Observability, Builds, and Audit Logs.

MCP is optional. This skill works with MCP (auto), Wrangler CLI, or Dashboard. See BACKENDS.md for execution options.

Permission Tiers

Security Rules

Read Operations

1. Define scope first — account / worker / resource ID
2. No account set? — List accounts first, then set active
3. Evidence required — Conclusions must have logs/screenshots/audit records

Write Operations (Three-step Flow)

1. Plan: Read current state first (list/get)
2. Confirm: Output precise change (name/ID/impact), await user confirmation
3. Execute: create/delete/update
4. Verify: audit logs + observability confirm no new errors

Prohibited Actions

• ❌ Execute create/delete/update without confirmation
• ❌ Delete production resources (unless user explicitly says "delete production xxx")
• ❌ Use Super Admin privileges in non-isolated environments
• ❌ Use container sandbox as persistent environment

Operation Categories

Diagnose Tier (Read-only)

Change Tier (Write Operations)

⚠️ = Requires confirmation

Super Admin Tier (Container Sandbox)

Temporary container for isolated tasks (~10 min lifecycle):

• Initialize, execute commands, read/write/delete files
• Use for: running tests, reproducing issues, parsing data
• NOT for: persistent state, production workloads

Common Workflows

Troubleshooting Flow

1. Clarify symptoms → worker name / time range / error type
2. Query observability to pull logs/metrics
3. If build-related → get build logs
4. If page-related → take screenshot to reproduce
5. Trace changes → pull audit logs
6. Summarize: root cause + evidence + fix recommendations

Resource Management Flow

1. List accounts → set active account
2. List resources (KV / R2 / D1)
3. Plan changes → present to user
4. Execute after confirmation
5. Verify: audit logs + observability shows no errors

Output Format

• Language: English
• Structure: Conclusion → Key data/evidence → Tool call summary → Next steps
• Write operations: Must clearly list operations and impact scope

Example:

✅ Investigation complete: worker `api-gateway` experienced 5xx spike between 18:00-18:30

Root cause: New code deployed threw TypeError when processing /v2/users
Evidence:
- Logs: 18:02 first occurrence of "Cannot read property 'id' of undefined"
- Audit: 18:00 user dev@example.com deployed new version
- Metrics: error_rate jumped from 0.1% to 12%

Recommendation: Roll back to previous version, or fix /v2/users handler

File-based Pipeline

When integrating into multi-step workflows:

runs/<workflow>/active/<run_id>/
├── proposal.md                # Symptoms/objectives
├── context.json               # Account/worker/resource/time_range
├── tasks.md                   # Checklist + approval gate
├── evidence/observability.md
├── evidence/audit.md
├── evidence/screenshots/
├── evidence/change-plan.md    # Write operations written here first
├── evidence/report.md         # Conclusion + evidence + next steps
└── logs/events.jsonl          # Optional tool call summary

Error Handling

Related Files

• BACKENDS.md — Execution options (MCP/CLI/Dashboard)
• SETUP.md — MCP configuration (optional)
• scenarios.md — 20 real-world scenario examples