SPM
Workflows

Daily and weekly pipelines that compound.

Each workflow is a skill — invokable, repeatable, logged. Some are live today, some are archived for resurrection, some are landing soon.

Invokable

One command. No memory of yesterday needed.

Every workflow ships as a Claude Code skill. Type the slash command, get the artifact. No "where did we leave off?" questions.

Repeatable

Same shape, same gate, every time.

Output structure is fixed. New PAD on Monday looks like Tuesday's PAD. Predictability is what makes them compound — agents can read their own history.

Logged

Every run leaves a receipt.

MongoDB + filesystem mirror. You can rerun, replay, or audit any past run. Nothing important lives in a chat log nobody reads.

Getting started

Running your first workflow.

Five steps from zero to a scheduled daily workflow. PAD is the entry point — start there.

  1. 01

    Install Claude Code

    npm install -g @anthropic-ai/claude-code — or open the VS Code extension. Skills live at ~/.claude/skills/ on your machine.

  2. 02

    Clone the PAD skill

    The simplest entry point. Copy the ~/.claude/skills/pad/ directory to your machine. Read SKILL.md — it is a plain markdown file, nothing more.

  3. 03

    Set your context

    Edit the memory/ folder inside the skill to describe your project. That is the only config — the skill uses it on every run.

  4. 04

    Run /pad

    Type /pad in Claude Code. First run takes ~60 seconds. Output is a dated markdown page in your gallery folder. Read it — the shape is the spec.

  5. 05

    Schedule it

    Once you like the shape, schedule /pad via /schedule or a system cron. From here it runs daily without you, building compounding context.

Archived · resurrectable

Built once, dormant now.

These earned their keep, then made way. The skill files still ship — one POM can bring any of them back online.

How they compose

One loop, many feeds.

Every daily workflow is a feed into the master AI PM loop. PAD gives you context. LAD gives you a marketing artifact. Appadday gives you a daily shipped surface. Together they keep the loop fed without burning the operator.

Context in
  • · PAD
  • · LAD
  • · BAD (soon)
Build / ship
  • · Appadday (soon)
  • · Coding playbook
Media out
  • · IAD
  • · VAD
  • · AAD
Loop close
  • · UAT triage
  • · Prod telemetry
  • · Vx+1 brief
A day in the loop

What a working day actually looks like.

The loops aren't a Gantt chart. They're a cadence — three windows a day where the agent is on, plus async windows where the operator is. Below is the canonical rhythm that runs the 13-site fleet.

Morning · 30 min
PAD

Agent pre-reads overnight CI runs, support inbox, and PR queue. Operator picks the day's three PBIs from the briefed shortlist. No standup needed — the brief is the standup.

Midday · 90 min
AAD + IAD + VAD

Media loops run in parallel: image generation, audio summaries, video drafts. Operator reviews the daily LAD output (one ranked top-100 list) and triages flagged items in the BO.

End of day · 20 min
LAD + Sprint close

Daily LAD is upserted to AIDB and the gallery. Velocity log auto-fills from the day's merged PRs. Tomorrow's PBIs get seeded into the brief — the agent picks them up before the operator wakes.

The shape of the win

Three sprint-close numbers from the fleet.

22d → 6d

Cycle time, idea → prod (e-commerce engagement)

400 → 80

Backlog tickets, eight weeks (fintech engagement)

Week 10 / 16

Early delivery on a 16-week API contract (SaaS engagement)

Full case-anonymised stories on the case studies page.