Skip to main content

PTrends Process Shims

Operator-facing reference for the process-shim model ratified in Plan #30. Audience: agents and operators who invoke or author local surface helpers. Governing SPEC: SPEC-138 v0.1 (model + authority boundary), SPEC-137 v0.1 (generation contract).

What PTrends Means

PTrends is the neutral label Prism uses for prevailing workflow-packaging trends that are translated into Prism-native governance language. It is a source label only. No committed artifact names an external source, tool, repository, project, or author as provenance for the process-shim model.

What a Process Shim Is

A process shim is a thin advisory pointer. It helps an agent or operator discover, invoke, format, or record use of a Prism process body. It is not the process body itself and it is not authority. Shims live on local surfaces — editor helpers, skills, slash commands, and similar host-local conveniences. They are generated and managed per SPEC-137.

What a Process Shim Is Not

A shim mayA shim may not
Reference a stable process id or lookup keyDuplicate or restate full process body text
Carry a concise label and one-line purposeDefine governance rules
Carry trigger hints for auto-invocationGrant approval, merge, deploy, secret, or RTE authority
Reference output and evidence shapesOverride a Ring rule, committed governance file, or ratified contract
Carry routing hints for the owning laneRead or write local editor memory as a source of truth
Reference SPEC-137 for generation mechanicsAct as a registry that can drift from Prism
When a shim conflicts with a Ring rule, committed governance file, ratified Prism SPEC or ADR, role or persona contract, or a live Prism process body, the shim loses.

Authority Model

Authority is fixed and does not shift when a shim is installed or invoked:
  1. Ring 0 — native BIOS files (AGENTS.md / CLAUDE.md) for universal non-negotiable session invariants.
  2. Ring 1–3$PRISM_ROOT/PRISM.md, $PRISM_ROOT/Orgs/$PRISM_ORG/ORG.md, and $PRISM_PROJECT_DIR/PROJECT.md for committed governance.
  3. Ratified Prism artifacts — SPECs, ADRs, role contracts, persona contracts, and process bodies.
  4. Runtime Prism servicesprism_start, signals, semantic recall, process lookup, and Team Memory as live operational state and evidence.
  5. Local shims — invocation and formatting conveniences only.
A shim sits at layer 5. It cannot elevate itself or grant authority that its layer does not have.

Two-Layer Model

The process-shim model is split into two SPEC layers so that the principle and the mechanics stay cleanly separated:
LayerSPECOwns
ModelSPEC-138Principle, authority boundary, packaging stance, PO outer-loop expectation, governance acceptance criteria
Generation contractSPEC-137Generated-shim metadata, process-body fetch and verification, stale/missing-body behavior, trigger confidence, missing_context_policy, invocation evidence, tests
When you need to understand what a shim is allowed to be, read SPEC-138. When you need to understand how a shim is generated, fetched, verified, or tested at runtime, read SPEC-137.

Agent-Driven Operating Model

Prism is agent-driven. Process shims do not turn it into a human-driven command system. Humans may invoke a shim as a shortcut. Agents — especially Product Owner agents — must invoke the applicable Prism process automatically when the lifecycle context requires it and the SPEC-137 trigger gate passes. An agent must not wait for a human to type a command when the current work clearly requires planning, assignment, review, merge, deploy, verification, incident routing, or learning closure. The agent must not rely on broad keyword matching as proof that a process should run. Trigger confidence and missing-context behavior are governed by SPEC-137.

Product Owner Outer Loop

The Product Owner archetype owns the outer development lifecycle loop:
Think → Plan → Assign → Build → Review → Merge → Deploy → Verify → Learn
Each stage iterates until its exit evidence is satisfied. The Product Owner must:
  • Read or refresh the outer-loop process at bootstrap and at major stage transitions when the applicable trigger gate passes.
  • Detect stalled stages and assign ownable work immediately.
  • Drive frequent safe merge, deploy, and smoke-verify cadence.
  • Close each loop iteration with evidence and learning records.
The Product Owner must not wait for a human slash command when the lifecycle context clearly requires action and the trigger gate passes.

How Agents and Humans Use Shims

Agents auto-invoke when trigger conditions are met (per SPEC-137). The shim points to the process body; the agent fetches and applies it. The shim does not replace the process body or change what authority the agent has. Humans may invoke a shim via a slash command, editor skill, or helper surface as a shortcut to discovering and running a process. The same authority rules apply — the shim is a pointer, not a grant. In both cases, the Prism process body fetched by the shim is the authority. The shim that pointed to it is not.

Acceptance Gate (Docs Lane)

Per SPEC-138 §Acceptance Criteria, the docs lane confirms:
  • Operator-facing docs explain PTrends and process shims without external source names.
  • No doc implies humans must drive every lifecycle step.
  • Shims are described as thin advisory pointers that fetch and apply Prism process bodies, not as authority.
  • Ring/system contracts and Prism process bodies remain the source of truth.
This page satisfies criterion 13. Desiree, docs lane.

References

  • SPEC-138 v0.1 — PTrends Process-Shim Model (authority boundary + governance).
  • SPEC-137 v0.1 — Process-Shim Generation Contract (mechanics + tests).
  • SPEC-078 — Consensus-First Parallelism and Method Fragments.
  • SPEC-105 — Agent Contracts.
  • SPEC-113 — Bootstrap Path Resolution and Scaffold Completeness.
  • SPEC-130 — Team Memory Events.
  • Plan #30.
Last modified on May 23, 2026