011matthias

system-digest

Generates structured system digests (overview, changelog, client report) with internal and client-facing templates. Scans live system state and delivers via terminal, file, or email. Use when running /system-digest or when user wants a system overview, progress report, or client update.

011matthias 0 Updated 3mo ago

Resources

1
GitHub

Install

npx skillscat add 011matthias/agentic-ops1-01/system-digest

Install via the SkillsCat registry.

SKILL.md

System Digest

Scans the live Agentic Ops system and generates structured digests for system awareness, client updates, or progress reporting.

Quick Start

/system-digest --overview              # What does this system do?
/system-digest --changes --since DATE  # What changed recently?
/system-digest --client NAME           # Client-specific report

Available Operations

Operation Description Module
System Overview Full inventory + architecture + per-client summary OVERVIEW-TEMPLATE.md
Changelog Recent changes to primitives, clients, learnings CHANGELOG-TEMPLATE.md
Client Report Client-specific automations, status, configurability CLIENT-TEMPLATE.md
Email Delivery Send digest via Resend API DELIVERY.md

Process

Step 1: Parse Mode

Determine mode from arguments:

  • --overview (default): full system inventory
  • --changes [--since DATE]: recent changes (default: 7 days)
  • --client NAME: client-specific report

Determine template:

  • Internal (default): technical, for system operator
  • --client-facing: value-focused, for clients

Step 2: Scan System State

Read the following to build a live inventory:

Primitives:

  • ls .claude/rules/*.md → count + names
  • ls .claude/skills/*/SKILL.md → count + read first description line from each
  • ls .claude/agents/*.md → count + names
  • ls .claude/commands/*.md → count + names

Rules budget:

  • Count total lines across .claude/rules/*.md (budget: 250)

Client status:

  • For each client in workspace/clients/*/:
    • Read workspace/clients/{client}/specs/automation-status.yaml if exists
    • Read infrastructure.yaml if exists (for orchestrator type)
    • Count specs in specs/ subdirectories

Skill modules:

  • For each skill, count files in modules/ subdirectory

Step 3: Gather Changes (if --changes mode)

  • Read docs/checkpoints/ for checkpoint files newer than --since date
  • Read recent entries from MEMORY.md (scan for dates >= --since)
  • Check git log: git log --since=DATE --name-only -- .claude/ for changed primitives
  • Diff automation-status.yaml against previous digest if available

Step 4: Apply Template

Load the appropriate template module:

Each template has two variants:

  • Internal: technical details, skill names, module counts, architecture
  • Client-facing: automation descriptions, status badges, value statements

Step 5: Render

Generate the digest as markdown. If --html or --email is requested, convert to styled HTML.

Step 6: Deliver

  • Terminal (default): print the markdown output directly
  • --file: save to docs/digests/YYYY-MM-DD-{mode}.md
  • --html: save styled HTML to docs/digests/YYYY-MM-DD-{mode}.html
  • --email ADDRESS: follow DELIVERY.md instructions

Modules