patrol-human-edits

star 1

Reviews recent human or agent edits to wiki pages for policy, provenance, and citation risk before any lane downstream of change-patrol can open. Use when change-patrol needs a diff-based risk classification for changed wiki content.

wryenmeek By wryenmeek schedule Updated 4/25/2026

name: patrol-human-edits description: Reviews recent human or agent edits to wiki pages for policy, provenance, and citation risk before any lane downstream of change-patrol can open. Use when change-patrol needs a diff-based risk classification for changed wiki content.

Patrol Human Edits

Overview

This skill documents the edit-patrol step for the change-patrol persona. It inspects a diff (or set of changed pages) for policy risk — citation removals, destructive edits, NPOV violations, and provenance gaps — and produces a risk classification. The patrol result gates downstream maintenance and quality lanes; it does not remediate content directly.

Doc-only workflow. No logic/ dir is introduced.

Classification

  • Mode: Doc-only workflow
  • MVP status: Active
  • Execution boundary: Read-only diff assessment only. No content change or remediation is performed.

When to Use

  • A human editor or external agent has committed changes to wiki/**
  • change-patrol must classify the risk before downstream lanes open
  • An operator suspects a destructive or policy-violating edit has occurred
  • A CI/CD pipeline needs a governed diff review before any write-capable step proceeds

Contract

  • Input: a diff or set of changed page paths, along with the before/after content
  • Output: a structured risk classification per change, including the risk level (low/medium/high), affected policy areas, and recommended next lane
  • Handoff: high-risk findings route to log-patrol-incident; low-risk findings allow the normal downstream lane to proceed
  • AFK safety net: for changes classified as AFK (per ADR-014), apply stricter review thresholds — flag any citation changes, new claims, topology changes, or entity references that should not appear in an AFK-classified task

Assertions

  • No content is modified by this skill
  • Citation removals, deleted provenance sections, and NPOV violations are always classified as high risk
  • The patrol result is the required input for any downstream maintenance or quality lane that follows a human edit
  • Missing or ambiguous diff input fails closed
  • Changes classified as AFK are subject to stricter review: citation modifications, new claim introductions, topology changes, and new entity references are always flagged as potential misclassifications requiring reclassification to HITL

References

  • AGENTS.md
  • docs/architecture.md
  • raw/processed/SPEC.md
  • .github/skills/policy-diff-review/SKILL.md
  • .github/skills/log-patrol-incident/SKILL.md
  • .github/agents/change-patrol.md
  • docs/decisions/ADR-014-hitl-afk-work-classification.md
Install via CLI
npx skills add https://github.com/wryenmeek/knowledgebase --skill patrol-human-edits
Repository Details
star Stars 1
call_split Forks 0
navigation Branch main
article Path SKILL.md
More from Creator