AGENT SKILL

incident-response

Investigate and triage runtime incidents involving the Adobe Dispatcher Apache HTTP Server module and related HTTPD configuration in AEM as a Cloud Service environments only, using runtime-backed MCP evidence.

License 许可
Apache-2.0
所属仓库
kalash33/kstack
最近更新
2026-05-20
Stars
★ 0
Forks
0

01 前言 Front Matter

name
incident-response
license
Apache-2.0
metadata
{ "mcp-tool-contract": "core-7-tools" }
description
Investigate and triage runtime incidents involving the Adobe Dispatcher Apache HTTP Server module and related HTTPD configuration in AEM as a Cloud Service environments only, using runtime-backed MCP evidence.
allowed-tools
validate lint sdk trace_request inspect_cache monitor_metrics tail_logs
compatibility
Requires Dispatcher MCP configured for cloud variant (`AEM_DEPLOYMENT_MODE=cloud`).

02 资源文件 17

03 SKILL.md 正文

Dispatcher Runtime Incident Response (Cloud)

Investigate cloud runtime incidents involving the Adobe Dispatcher Apache HTTP Server module and related HTTPD configuration with an evidence-first workflow.

Variant Scope

  • This skill is cloud-service-only.
  • Scope is fixed by this skill directory; do not ask the user to choose deployment variant.

MCP Tool Contract

Use only these Dispatcher MCP tools:

  • monitor_metrics
  • tail_logs
  • trace_request
  • inspect_cache
  • validate
  • lint
  • sdk

Workflow

  1. Quantify impact (monitor_metrics).
  2. Classify whether issue belongs to Dispatcher layer or CDN/edge layer using cloud guardrails, including reserved probe-path behavior.
  3. Gather logs and traces (tail_logs, trace_request). To get the full log sequence for a specific request: use tail_logs to obtain entries (some include pid and tid from dispatcher debug lines), then call trace_request(pid=..., tid=...) to retrieve all log lines for that request. See mcp-tool-orchestration.md § Trace by pid:tid.
  4. Inspect cache behavior (inspect_cache).
  5. Correlate with static checks (validate, lint, sdk).
  6. Return containment + remediation + rollback.

Verification Scope Selection

Use shared references to select incident evidence depth:

Output Contract

Always return:

  • incident summary (symptom, impact window, blast radius)
  • evidence table (tool, input, finding)
  • selected test IDs and outcomes
  • probable cause + confidence level
  • containment, durable remediation, and rollback
  • open risks and missing evidence

Guardrails

  • Distinguish observations from inference.
  • Do not claim root-cause certainty without corroborating evidence.
  • Do not claim runtime verification if container/runtime prerequisites were missing.
  • Keep fixes reversible unless user explicitly requests high-risk change.
  • For cloud incidents, explicitly state when root cause is outside Dispatcher config scope (CDN/WAF/edge policy).

References

04 同仓库更多 Skills