adobe/config-authoring

Dispatcher Config Authoring (AMS)

Design minimal, deterministic changes for the Adobe Dispatcher Apache HTTP Server module and related HTTPD configuration in AEM 6.5/AMS, then verify with MCP evidence.

Variant Scope

• This skill is AMS-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:

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

Workflow

AMS Config Progress
- [ ] 1) Confirm scope and acceptance criteria; if the repo layout is unclear, normalize it with `repo-layout-workflows.md`
- [ ] 2) Apply AMS 6.5 guardrails for tier routing, variables, flush ACL, and immutable boundaries
- [ ] 3) Decompose target URLs (path/selectors/extension/suffix) and use that model for all URL-based rules—filters, cache rules, etc.—using `/path`, `/selectors`, `/extension`, `/suffix` or aligned globs (not raw `/url`) where applicable; then design complete section-level edits
- [ ] 4) Update config with least-privilege defaults (produce final merged section, not isolated rule snippets)
- [ ] 5) Run static checks: validate -> lint (deep/order-aware when filters changed)
- [ ] 6) Run SDK checks: `sdk({"action":"check-files","config_path":"<dispatcher src path>"})`, `sdk({"action":"diff-baseline","config_path":"<dispatcher src path>"})` as needed
- [ ] 7) Run runtime verification in host-path/log environment
- [ ] 8) Return diff, evidence table, risk/rollback, and citations

Verification Scope Selection

Use the shared references to select the minimum evidence set:

• mode-specific-verification-matrix.md
• test-case-catalog.md

Output Contract

Always include:

• files changed + intent
• exact insertion location and final merged section content for each edited block
• checks executed + evidence
• selected test IDs
• risk/rollback plan
• residual risks and next checks

Feature cross-boundary (Playbook G: auth_checker)

Permission-sensitive caching (/auth_checker) is end-to-end: it requires both Dispatcher config and an AEM servlet. When implementing Playbook G from scratch, create or verify the auth-check servlet in the project core bundle (path /bin/permissioncheck, HEAD/GET, 200 or 403) and allowlist it on publish; then add the /auth_checker block, filter allow for the endpoint, and /allowAuthorized "1" in /cache. See config-scenario-playbooks.md Playbook G and reference-snippets.md.

Guardrails

• Do not weaken deny-by-default security posture without explicit user approval.
• Do not claim runtime verification if host-path/log prerequisites were missing.
• Keep changes minimal and scoped.

References

• config-patterns.md
• config-scenario-playbooks.md – high-value development scenarios adapted from broader MCP surfaces to core-7 execution
• reference-snippets.md – reusable starter snippets for consistent config authoring
• validation-playbook.md
• mcp-tool-orchestration.md
• For runtime prompts and troubleshooting scenarios (tail_logs, trace_request, monitor_metrics, inspect_cache), use the incident-response skill reference runtime-prompts-and-troubleshooting-scenarios.md when incident triage is needed.
• quick-start-execution-path.md – single entry path for new users and broad requests
• repo-layout-workflows.md – map the user repo to the dispatcher src root and likely file families
• playbook-command-linkage.md – convert selected playbooks into exact MCP command chains
• ams-6-5-guardrails.md
• mode-specific-verification-matrix.md
• test-case-catalog.md
• change-risk-and-rollback-template.md
• public-doc-citation-rules.md
• public-docs-index.md
• core-7-tools-reference.md