PCC Agent ¶

PCC Agent is a production EAP application built to support PCC operational workflows on a governed agent runtime.

It encodes a high-volume operational workflow as a customer agent repo, registers it through enterprise-agent-platform-registry, and runs it under EAP's governance controls.

Why PCC Runs On EAP ¶

PCC work involves the same sequence of evidence gathering, dependency checks, and decision support across many alerts. PCC Agent turns that into reviewed skills running on a governed runtime.

Outcomes the platform model enables:

• less analyst toil during triage
• faster collection of relevant evidence
• more consistent decisions for repeated cases
• clear boundaries between low-risk automation and actions that need approval

How PCC Agent Maps To EAP ¶

Execution Flow ¶

sequenceDiagram
    participant Alert as Alert source
    participant Gateway as EAP gateway
    participant Runtime as EAP runtime (sandboxed)
    participant Skill as PCC skill
    participant Tools as Approved MCP
    participant Approval as Approval workflow
    participant Audit as Audit store

    Alert->>Gateway: Send alert context
    Gateway->>Gateway: Auth (MyID group or CI client_id)
    Gateway->>Runtime: Start sandboxed session
    Runtime->>Runtime: Inject platform guardrails
    Runtime->>Skill: Load approved PCC skill
    Skill->>Tools: Gather metrics, topology, incidents
    Tools-->>Skill: Evidence bundle
    Skill->>Runtime: Recommend next step
    Runtime->>Approval: Request approval if action is gated
    Approval-->>Runtime: Approve or reject
    Runtime->>Audit: Persist summary, evidence, token usage

Safety Model In Action ¶

PCC Agent inherits every EAP guarantee — it doesn't have to reimplement them:

• gather evidence before attempting any state-changing action
• stop when the case falls outside the declared procedure
• require approval for actions that can change production state
• emit a full execution trace for post-hoc review

Domain Scope ¶

The current PCC project covers domains such as R&S, OCP, CCE, and CDI. The runtime is shared; the skills and the approved MCP set are PCC-specific.

What A Good PCC Skill Produces ¶

Each approved PCC skill should return:

• normalized evidence
• a clear decision summary
• the recommended next action, or the executed action when approval and policy allow
• the reason the skill stopped, escalated, or requested approval

Reference For New Projects ¶

PCC Agent makes the things every EAP project must declare — dependencies, evidence, approval boundaries — explicit and reviewable, so new project owners can use its repo and registry layout as a reference.

When you build your own project, the shape is the same: a customer agent repo, an enterprise-agent-platform-registry YAML, and execution under EAP's runtime contract.