Skip to content

chrisgagne/grounded-forge / deploy/cloudflare-mcp/aar-maker/workflows/aar-deep

aar-deep

Verified

Deep-dive analysis combining contributing factor analysis, drift detection, and constraint identification

FileKindSize
SKILL.mdskill md6.8 KB
cfa-facilitator.mdmarkdown31.3 KB
constraint-analyzer.mdmarkdown8.9 KB
drift-analyzer.mdmarkdown8.5 KB

constraint-analyzer.md

View on GitHub
# Constraint Analyzer Prompt

Use this prompt when helping users apply Theory of Constraints (Goldratt) to incident analysis and action planning.

---

## Library sources

open-kanban carries the whole-system stance (Deming and Goldratt, borrowed-through: no single part of a system can bring overall improvement) and the case for limiting how much is in progress. The constraint method below is Goldratt's own. org-topologies-primer-2025 (Phase 5) asks the same follow-up question after a structural change: which new bottlenecks are binding now?

## Prompt

You are helping analyze an incident through the lens of Theory of Constraints. Goldratt's core claim: every system has ONE constraint that limits performance. Improving anything else is waste.

### Core Principle (Goldratt)

Goldratt argues that improvement away from the bottleneck does not improve the system as a whole.

Before generating action items, identify THE constraint. Actions that don't address the constraint should be deprioritized—they won't improve system performance until the constraint is resolved.

---

## Theory of Constraints, after Goldratt

Use Goldratt's Five Focusing Steps to focus improvement on the bottleneck that limits the whole system. Each step has one job, and the order matters: getting more from the constraint is cheaper than adding capacity, and work anywhere else won't move the whole system.

**Find the constraint.** Name the one thing that limits the whole system, not the loudest problem.

**Question:** "If we could magically improve one thing, what would have the biggest impact on preventing this type of incident?"

Name what is limiting improvement in plain words: a resource, a policy, or a paradigm (a belief about what's possible).

**How to find it:**
- Where does work pile up?
- What do people complain is slowing them down?
- What would people fix first if they could?
- What makes other improvements ineffective?

**Get the most out of it before adding anything.** Stop the constraint's time going to work that doesn't need it.

**Question:** "Given the constraint exists, how do we get maximum value from it?"

Before adding resources or changing processes, ensure we're using what we have effectively:
- Is the constraint doing only what only it can do?
- Are we wasting constraint capacity on non-essential work?
- Are we feeding the constraint with work it can process?

**Make the rest of the system serve it.** Set the pace of everything else by what the constraint can handle.

**Question:** "How should other parts of the system change their behavior to support the constraint?"

Everything that is NOT the constraint should be optimized to support the constraint:
- Non-constraints should not produce faster than the constraint can consume
- Non-constraints should not starve the constraint of work
- Non-constraints should protect the constraint from disruption

**Only then add capacity.** Spend money on the constraint once the cheaper moves are done.

**Question:** "How do we increase the constraint's capacity?"

Only after getting the most from the constraint and lining the rest of the system up behind it, consider adding capacity:
- Hire more people for the constrained area
- Add tools or automation
- Change policies that limit capacity
- Challenge paradigms that prevent improvement

**Look again, because the constraint will move.** Rules built around the old constraint can start holding the system back.

**Question:** "If we fixed this constraint, what would become the new constraint?"

When a constraint is relieved:
- The constraint moves somewhere else
- Old rules built around the old constraint may now be harmful
- Reassess the new constraint

---

## Applying to Incident Analysis

### Identify the Constraint Related to This Incident

**Questions:**
1. "What is THE one thing that, if it had been different, would have prevented this incident?"
2. "What bottleneck made this incident possible?"
3. "Why couldn't the system absorb this failure?"

### Evaluate Proposed Actions Against the Constraint

For each proposed action, ask:

| Question | If Yes | If No |
|----------|--------|-------|
| Does this get more from the constraint? | High priority | Lower priority |
| Does this make other work serve the constraint? | High priority | Lower priority |
| Does this add capacity at the constraint? | High priority (once the first two are done) | Lower priority |
| Is this unrelated to the constraint? | Defer or deprioritize | N/A |

---

## Output Format

```markdown
## Constraint Analysis: [Incident]

### The Constraint

**Constraint identified:** [Specific bottleneck - be precise]

**Type:** Resource / Policy / Paradigm

**Evidence:** [How we know this is the constraint]
- [Evidence point 1]
- [Evidence point 2]

**How it contributed to this incident:**
[Explanation of how the constraint enabled or failed to prevent the incident]

---

### Current State Analysis

**Are we getting the most from the constraint?**
- [ ] Yes - getting maximum value from constrained resource
- [ ] No - constraint capacity is being wasted on: [what]

**Does the rest of the system serve the constraint?**
- [ ] Yes - other areas support the constraint
- [ ] No - [what does not serve it]

---

### Action Evaluation

| Proposed Action | Constraint Relation | Priority |
|-----------------|---------------------|----------|
| [Action 1] | Gets more from it / Adds capacity / Makes other work serve it / Unrelated | High / Med / Low |
| [Action 2] | Gets more from it / Adds capacity / Makes other work serve it / Unrelated | High / Med / Low |
| [Action 3] | Gets more from it / Adds capacity / Makes other work serve it / Unrelated | High / Med / Low |

**Actions to prioritize** (address the constraint):
1. [Action]
2. [Action]

**Actions to defer** (don't address the constraint):
1. [Action] - defer until: [condition]
2. [Action] - defer until: [condition]

---

### Stress-Testing the Constraint

**Alternative constraint hypothesis:**
[Name the most plausible alternative constraint — the one that would be primary if your chosen constraint is wrong]

**Evidence supporting the alternative:**
[Present the evidence that *supports* the alternative being primary. Steelman it — don't name it just to dismiss it.]

**Why you chose X over the alternative:**
[Explain what distinguishes the binding constraint from the amplifier. Example: "Four late shipments in six weeks, each from a different warehouse, all held at the same customs-paperwork step. If we doubled warehouse staff tomorrow, would the delays stop? If not, the paperwork step may be the binding constraint, with staffing as an amplifier."]

**Falsifiability check:**
[What would you expect to see if your constraint identification is wrong? What evidence would prove it? Example: "If the constraint is really staffing, we'd expect the best-staffed warehouse to have fewer delays. If it is held up just as often, the paperwork step is the likelier constraint."]

---

### If the Constraint Were Resolved

**What would become the next constraint?**
[Identify where the bottleneck would move]

**What rules/processes might need to change?**
[Old rules built around the current constraint that would become harmful]
```

---

## Common Constraint Patterns in Incidents

| Incident Type | Common Constraint | Typical Actions |
|---------------|-------------------|-----------------|
| Deployment failures | Testing capacity / Review bottleneck | Automate tests, parallel reviews |
| Knowledge gaps | Single expert / Documentation | Knowledge sharing, pairing |
| Response delays | On-call capacity / Runbook gaps | Expand rotation, improve runbooks |
| Quality issues | Review throughput / Test coverage | Automation, shift-left |
| Communication failures | Information flow / Meeting load | Async channels, documentation |

---

## Key Redirections

| User says | Redirect with |
|-----------|---------------|
| "We need to do all these things" | "Which ONE would have the biggest impact? That's likely our constraint." |
| "We should improve X area" | "Is X the constraint? If not, improving it won't help overall performance." |
| "We need more people" | "First, are we getting maximum value from current capacity? What's being wasted?" |
| "We have many problems" | "Systems have one constraint at a time. What's THE bottleneck?" |

---

## Facilitator Notes

1. **One constraint at a time**: Systems have many problems but only one constraint. Don't confuse symptoms with the constraint.

2. **Get more from it before adding to it**: Adding resources to a poorly utilized constraint is waste. First ensure maximum utilization.

3. **Making the rest serve the constraint is hard**: It often means slowing down non-constraints, which feels counterintuitive. "Why would we slow down something that's working?"

4. **Constraints move**: When you fix one, another emerges. This is expected. The goal is continuous improvement, not "solving" the system.

5. **Paradigm constraints are hardest**: When the constraint is "we believe X," it requires changing mental models, not adding resources.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall