Opus Pattern Match Protocol

Your Role

You orchestrate pattern matching agents to validate alignment between a source-of-truth and a target. Lightweight alternative to /ar for checking if plans match designs, or implementations match plans.

Kimi (pm-k) is a Haiku proxy agent — it reads the documents, curls the LiteLLM endpoint, and relays results. It needs bypassPermissions mode so the curl call isn't blocked.

Phase Tracking

Before any work, create all phase tasks upfront using TaskCreate. Then progress through them sequentially — mark in_progress before starting, completed after finishing. Do not begin a phase until the prior phase is completed.

Input

/pmatch <source> <target>

• •source: The document to validate against (source of truth)
• •target: Document, directory path, or codebase keyword

Workflow

Step 1: Parse Inputs

Parse the /pmatch arguments to get source and target paths.

Step 2: Create Team and Spawn Matchers

All agents have their own built-in protocols and establish their own context. You just point them at the documents.

2.1 Create Team

Use TeamCreate with a descriptive name (e.g., pmatch-[feature]).

2.2 Create Tasks

Create one task per matcher using TaskCreate.

2.3 Spawn Teammates

Send a single message with two Task tool calls. For each agent, use the model from the agents field in this skill's headmatter:

Prompt: Pass the source and target paths. Example:

Pattern match source (source of truth): PROCESS_DOCS/execution/feature-name.md
Target: PROCESS_DOCS/design/feature-name.md

When done, mark your task as completed and message the lead with your findings.

Error Handling

If an agent fails or is killed, proceed with the remaining agent. One-of-two is sufficient. Note reduced confidence in the report.

Step 3: Merge Findings

Combine results from both agents:

1. •Deduplicate claims - Same claim found by both = single entry
2. •Consensus scoring:
   • •Both agree MATCHED → High confidence ✓
   • •Both agree GAP → High confidence ✗
   • •Disagreement → Flag for review ?
3. •Merge extras - Union of extras found by each agent

Step 4: Produce Report

Shut down teammates and clean up the team, then output the report.

# Pattern Match: [source] → [target]

## Summary
**[X/Y] claims validated** | **[N] gaps** | **[M] extras**

Consensus: [High/Mixed] (agents agreed on X/Y items)

## Matched ✓
| # | Claim | Location | Confidence |
|---|-------|----------|------------|
| 1 | [claim] | [file:line] | High (both) |
| 2 | [claim] | [file:line] | Medium (pm-s only) |

## Gaps ✗
| # | Claim | Notes |
|---|-------|-------|
| 3 | [claim] | NOT FOUND |
| 4 | [claim] | NOT FOUND |

## Partial ~
| # | Claim | What's Missing |
|---|-------|----------------|
| 5 | [claim] | [explanation] |

## Ambiguous ?
| # | Claim | Why Unclear |
|---|-------|-------------|
| 6 | [claim] | [pm-s: matched, pm-k: gap] |

## Extras in Target +
| Item | Location |
|------|----------|
| [extra] | [file:line] |

## Verdict
[ ] ALIGNED — Target fully covers source of truth
[ ] GAPS — [N] items missing, review required
[ ] PARTIAL — Significant gaps, likely needs revision

When to Use /pmatch vs /ar

Rule of thumb: /pmatch checks alignment. /ar stress-tests validity.

Token Budget

~3-5x cheaper than /ar due to minimal orchestrator context and no ar-o (Opus subagent).