CM-Stack

An AI coding-agent playbook built on The Five Pillars -- Clear Measure's framework for leading effective custom software teams.

"Software projects don't fail because of bad code. They fail because of missing leadership." -- Jeffrey Palermo, The Five Pillars: Leadership for Effective Custom Software, Chairman, Clear Measure, Inc.

Why CM-Stack exists

Clear Measure has spent two decades helping companies keep custom .NET software projects on track -- visible, deliverable, and built on engineering practices that actually hold up in production. Jeffrey Palermo distilled that experience into The Five Pillars: Create Clarity, Establish Quality, Achieve Stability, Increase Speed, and Optimize the Team.

CM-Stack turns those Five Pillars into slash-command skills for your AI coding agent (GitHub Copilot CLI, Claude Code, or any Markdown-skill-compatible agent). Instead of one blank prompt, you get five specialists -- each one grounded in a pillar of the book -- plus a scorecard that reports where your project actually stands.

This isn't a rewrite of the book. It's an operating system for applying it, every day, inside the tools your engineers already use.

The Five Pillars, as skills

Pillar	Skill	What it does
1. Create Clarity	/clarity	Checks a project, feature, or team against its stated vision, strategy, and measurement plan. Flags "running blind."
2. Establish Quality	/quality	Reviews code, PRs, or a defect backlog for escaped-defect risk and process controls -- before defects reach users.
3. Achieve Stability	/stability	Audits production readiness, incident history, and recovery process. Protects the "last mile to ROI."
4. Increase Speed	/speed	Analyzes work-item flow, decomposition, and throughput -- replacing estimates with counted, forecasted delivery.
5. Optimize the Team	/team	Assesses team health (forming/storming/norming/performing) and recommends how to strengthen the players you have.
All Five	/scorecard	Runs all five pillar checks and produces a single leadership scorecard for the project, with trend vs. the prior run.
Values	/values-check	Pressure-tests a decision or plan against Clear Measure's core values before you commit to it.

Fast-path companion skills

Narrower, single-scenario skills that map to one pillar without running the full check:

Pillar	Skill	What it does
3. Achieve Stability	/postmortem	Writes a single-incident postmortem right after production incident is resolved.
4. Increase Speed	/decompose	Breaks one epic or feature into day-sized work items.
5. Optimize the Team	/onboarding	Builds a new-hire ramp-up plan and flags who's set up to fail by circumstance.

See docs/FIVE_PILLARS.md for the full framework and skills/ for each skill's definition.

Who this is for

• Software executives and tech leads at Clear Measure who want AI-assisted oversight, not just AI-assisted typing
• Engineering managers standing up a new team or rescuing a struggling project
• Delivery leads who need a repeatable way to show clients their project is on track

Quick start

1. Install -- clone this repo into your agent's skills directory:

   git clone https://github.com/clearmeasure/cm-stack.git $env:USERPROFILE\.copilot\skills\cm-stack

   (Swap the destination for your agent's skill path -- e.g. ~/.claude/skills/cm-stack for Claude Code.) For Cursor, Windsurf, plugin-marketplace installs, or any other harness, see the full Installation Guide.

2. Run /clarity on your current project to see if your team can answer "why" in one sentence.

3. Run /quality on your last five merged PRs to see how many defects would have escaped.

4. Run /scorecard to get the full five-pillar picture in one pass.

5. Run /values-check before your next big decision.

No project handy yet? See examples/atlas-api/ -- a fictional sample project with fixtures for every skill, so you can try /clarity, /quality, /stability, /postmortem, /speed, /decompose, /team, /onboarding, /scorecard, and /values-check before pointing any of them at real data. The Five Pillars apply beyond software delivery too -- see examples/meridian-advisory/ for the same walkthrough against a fictional professional-services/consulting engagement instead of a codebase.

Walkthrough: demo with examples/atlas-api/

examples/atlas-api/ is a fictional project (fake PRs, defects, incidents, throughput data, team notes, a client proposal, and prior scorecard history) with a deliberate finding baked into every fixture, so a skill run against it always has something concrete to report. Try it end to end:

1. Install CM-Stack (see Quick start above), then open a session inside this repo (or copy examples/atlas-api/ into a project where your agent is already running).

2. Run /clarity on the vision doc:

   "Run /clarity on examples/atlas-api/vision.md"

   Expect a Yellow/Red/Green mix: Vision Yellow (2 of 3 team members agree), Strategy Red (nothing written past this quarter), Measurement Green (a real weekly scorecard exists). The skill should flag Strategy as the biggest gap and draft a one-sentence vision statement.

3. Run /quality on the PR and defect data:

   "Run /quality on examples/atlas-api/prs.json and defects.json"

   Expect all 3 escaped defects traced back to the 3 hotfix/* PRs that merged with zero reviews -- and a recommendation to require review on hotfix/* branches, since it's the single control that would have caught every defect in the sample.

4. Run /postmortem on the unresolved incident:

   "Run /postmortem on incident 4 in examples/atlas-api/incidents.md"

   Expect a timeline, root cause (unbounded retry loop exhausting the DB connection pool), and a callout that this is a repeat of Incident 1 -- a signal the April fix patched a symptom, not the root cause.

5. Run /decompose on the oversized epic:

   "Run /decompose on the bulk-export epic in examples/atlas-api/throughput.csv"

   Expect the epic split into day-sized sub-items with one piece flagged as cuttable scope.

6. Run /onboarding on the new hire:

   "Run /onboarding for engineer-c in examples/atlas-api/team-notes.md"

   Expect a flag that the tech lead is already overloaded (3 of 4 incident resolutions) and a recommendation to assign a different onboarding buddy.

7. Run /values-check on the billing proposal:

   "Run /values-check on examples/atlas-api/client-proposal.md"

   Expect flags on Serving Others First, Building Relationships, and Authenticity -- billing a defect fix as a feature favors short-term revenue over client trust.

8. Run /scorecard for the full picture, with trend:

   "Run /scorecard on examples/atlas-api/"

   Expect all five pillar scores plus a Trend column comparing this run to the two prior runs already recorded in examples/atlas-api/scorecard-history.jsonl.

Compare what you get back to the "Example output" section in each skill's skills/<name>/SKILL.md -- the fixtures were written to match, so the shape of the report (not necessarily every word) should look familiar.

Built on Clear Measure's values

CM-Stack's skills are written to reflect how Clear Measure works, not just what it measures:

• Serving Others First -- every skill's output is written for the person who has to act on it, not to make the agent look smart
• Building Relationships -- /scorecard output is meant to be shared with clients and teams, not hidden in a terminal
• Excellence -- skills favor counted facts (defects, throughput, incidents) over vibes
• Authenticity -- a red scorecard stays red until the underlying pillar improves
• Empowered Employees -- skills coach the team to self-diagnose, not just report to leadership
• Trusted Partner -- the same rigor Clear Measure brings to client engagements is baked into every skill

Read the full list at docs/VALUES.md or clearmeasure.com/about-us/our-values.

Learn more

• The book: The Five Pillars: Leadership for Effective Custom Software by Jeffrey Palermo -- Programmer, Chief Architect, Software Executive, 20-time Microsoft MVP, Chairman of Clear Measure, Inc.
• The company: clearmeasure.com -- Empowering Software Delivery
• This repo: fork it, adapt the skills to your stack, and make it your team's own operating system for shipping custom software that doesn't fail.

License

MIT -- see LICENSE. The Five Pillars framework and name are (c) Jeffrey Palermo / Clear Measure, Inc.; this repo is an independent, non-official companion toolkit and is not a reproduction of the book itself.