prism-mcp-orchestrator

name	prism-mcp-orchestrator
description	Auto-use for context-heavy prompts (logs, CI failures, Kubernetes issues, runbooks, docs lookup, PR triage, codegen subtasks). Decide when to delegate via Prism MCP and execute the required call order — list_agents, list_resources, get_resource, optional list_prompts/get_prompt, run_agent, then parent synthesis.
compatibility	Requires Prism MCP server availability, local repository context, and an MCP-capable host editor/agent.
metadata	{"prism-agents":"orchestrator parent-model"}

Prism MCP orchestrator skill

Use this skill for orchestration only (not domain diagnosis). If the task is evidence-heavy or specialist-shaped, delegate via Prism MCP.

Activation rule

Activate when prompts mention things like: CI failures, PR blockers, Kubernetes incidents, Argo drift/workflow failures, release notes/docs lookup, large logs/JSON, or "delegate/run_agent/MCP".

Do not activate for tiny single-step tasks where delegation overhead is not worth it.

Fail-closed evidence gate

Before delegation, check for minimum identifiers:

• GitHub: repo + (PR, run URL, or run ID)
• Kubernetes: context + namespace + workload (deployment/statefulset/pod)
• Release notes/docs: product/library + version range or target release

If required identifiers are missing, do not run specialists yet. Ask the parent LLM to request only missing fields.

First-party skill precedence

When a subtask targets a specific platform/tool/vendor, prefer that vendor's first-party skills/docs before community or generic sources.

Examples:

• LaunchDarkly tasks -> prefer LaunchDarkly first-party skills (for example from the LaunchDarkly ai-tooling/skills source) and official docs.
• Cloud/provider SDK tasks -> prefer official provider skills/docs first.

If a first-party source exists but is not available in the current environment, callback to parent with insufficient_evidence and request permission/inputs to access the first-party source.

Required sequence (must follow)

1. Call list_agents first.
2. Call list_resources, then get_resource for:
   • prism://resource/tooling/run_agent
   • prism://resource/tooling/orchestration-guide
   • prism://resource/agent/<agent_id>/constitution (for selected agents)
3. Optional: list_prompts + get_prompt for payload templates.
4. Call run_agent with bounded task + valid skill_names.
5. Synthesize specialist outputs in parent model.

If evidence is insufficient at any step, callback to parent with:

{
  "status": "insufficient_evidence",
  "needs_from_parent": [
    "repo owner/name",
    "PR URL or workflow run URL",
    "namespace/deployment"
  ],
  "reason": "Identifiers required for evidence-backed delegation are missing."
}

run_agent shaping

• Keep tasks narrow and evidence-oriented.
• Attach only relevant evidence per specialist.
• Prefer multiple focused calls over one mixed-domain call.
• Keep final judgment/tradeoffs/recommendation in parent synthesis.
• For tool-specific tasks, include first-party source preference in the task text so specialists ground findings in vendor-authored guidance.

Minimal call template

{
  "agent_id": "github-cli",
  "skill_names": ["gh-pr-triage"],
  "task": "Repository owner/repo. Triage PR #42 for merge blockers with evidence.",
  "format": "json"
}

Parent output contract

• Concise synthesis (no raw dumps)
• Ordered findings with evidence
• Confidence + concrete next action
• If evidence is missing: return insufficient_evidence with required follow-up fields, not speculative findings