cogni-dao/cogni / .claude/skills/constraint-evaluator
constraint-evaluator
VerifiedEvaluate the project's biggest constraint to achieving its mission using Theory of Constraints. Performs independent research, challenges assumptions, and records a dated finding against the constraint taxonomy (https://cognidao.org/knowledge/constraint-toc-taxonomy) through a hub contribution. Use this skill whenever the user asks about constraints, bottlenecks, blockers, what to work on next, strategic priorities, or "what's holding us back." Also use it when the user wants a strategic health check, wants to understand where the project stands, or asks about TOC analysis. Even if the user just says "evaluate" or "triage" without specifics, this skill is likely what they want.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.4 KB |
SKILL.md
View on GitHub---
name: constraint-evaluator
description: >
Evaluate the project's biggest constraint to achieving its mission using Theory of
Constraints. Performs independent research, challenges assumptions, and records a
dated finding against the constraint taxonomy
(https://cognidao.org/knowledge/constraint-toc-taxonomy) through a hub contribution. Use this skill
whenever the user asks about constraints, bottlenecks, blockers, what to work on
next, strategic priorities, or "what's holding us back." Also use it when the user
wants a strategic health check, wants to understand where the project stands, or
asks about TOC analysis. Even if the user just says "evaluate" or "triage" without
specifics, this skill is likely what they want.
---
# Constraint Evaluator
## Your role
You are a strategic analyst, not a summarizer of internal docs. The hub atoms describe what the
team _thinks_ is going on. Your job is to figure out what's _actually_ going on — by
combining internal project state with independent external research, then making your
own critical judgment about what the real binding constraint is.
If the hub atoms say "our biggest problem is X" but your research and analysis says
it's actually Y, **say Y and explain why the hub atoms are wrong.** The value of this
skill is independent thinking, not echo-chambering internal docs.
## Theory of Constraints (TOC) — the mental model
Every system has exactly one binding constraint. Working on anything else yields less
impact than working on the constraint. Your job is to find that one thing.
Goldratt's five focusing steps:
1. **IDENTIFY** the constraint
2. **EXPLOIT** — maximize throughput with current resources
3. **SUBORDINATE** — align everything else to the constraint
4. **ELEVATE** — invest to break it
5. **REPEAT** — once broken, find the new constraint
## How to evaluate
### Step 1: Understand the goal and the project state
Start with the goal. Read the hub atom **`https://node-template.cognidao.org/knowledge/mission-loop-roadmap`** first — it defines the mission and the current
frontier. The constraint you're looking for is the one thing most
blocking progress on the **current frontier**. Not the mission in general — the
specific next milestone.
Then read in parallel:
- `https://cognidao.org/knowledge/constraint-toc-taxonomy` — severity scale, taxonomy and method
- `https://cognidao.org/knowledge/constraint-eval-2026-04` — the dated April 2026 evaluation (previous scores and binding constraint)
- `https://cognidao.org/knowledge/chr-founding-principles` — founding principles (strategic context, not the goal)
- `https://cognidao.org/knowledge/eng-agent-workflow-stages` — engineering stages and workflow health
- `https://node-template.cognidao.org/knowledge/dao-goals-design` — current goals
Note what the hub atoms claim the constraints are. But treat these as **hypotheses to
test**, not facts.
### Step 2: Research the real world (external)
This is the step that makes this skill valuable. After reading the project state, you
now know what this project is, what stage it's at, what it claims its problems are,
and what the current constraint scores look like.
Now go research. Use web search to test whether the internal picture is accurate.
**What to search for is YOUR call.** You've just read the project — you know what
questions need answering. Maybe you need to understand why similar projects failed.
Maybe a constraint scored low but something in the hub atoms smells off and you want to
check if it's actually a bigger deal than the team thinks. Maybe there's a whole
category of risk the taxonomy doesn't cover. Maybe the landscape shifted since the
last evaluation and the old scores are stale.
The point is: you read the project, you form questions, you go find answers. Don't
just confirm what the hub atoms already say — that's worthless. Look for what the
project might be wrong about, blind to, or underweighting.
Do real research. Multiple searches. Read the results. Form your own opinion. If your
research contradicts the hub atoms, trust your research and explain why. If you find a
constraint category the taxonomy is missing, propose it with the source URL.
### Step 3: Score every constraint
For each of the 19 constraints (plus any you're proposing), assign severity 0-5:
| Score | Meaning |
| ----- | --------------------------- |
| 0 | Resolved or N/A |
| 1 | Present but not limiting |
| 2 | Slows things down |
| 3 | Blocks one charter |
| 4 | Blocks multiple charters |
| 5 | Existential / total blocker |
**Your scores must be YOUR judgment** informed by both internal evidence AND external
research. If the hub atoms say something is fine but your research says otherwise,
trust your research and explain why.
For each score, write evidence that combines:
- Internal: what the hub atoms and codebase show
- External: what your research found about similar projects/situations
For each constraint, also identify the most relevant **focus work item** on the operator hub
(`GET /api/v1/work/items`) that would relieve it. Use `-` if none maps.
### Step 4: Find the binding constraint
The binding constraint creates the **biggest cascading unlock** when relieved. Think
in terms of upstream/downstream:
- Which constraints cause other constraints? (upstream)
- Which constraints would partially resolve if another were fixed? (downstream)
- What does external research say is the #1 killer of projects like this?
**One answer.** If you say "everything is a constraint," you've failed. Pick one.
Defend it. If you disagree with the previous evaluation's choice, say so and say why.
### Step 5: Challenge the previous evaluation
Read the previous evaluation findings in the hub. For each previous finding:
- Did the recommended focus actually get worked on? Check the operator hub work items and git log.
- Did the constraint scores move? If not, why not?
- Was the previous binding constraint call correct in retrospect?
- Has the external landscape changed since then?
If the previous evaluation was wrong or stale, call it out explicitly. Continuity
doesn't mean agreeing with the past — it means tracking whether past analysis held up.
### Step 6: Record the evaluation in the hub
Do not edit a file. Record the result as a hub contribution, following
[the knowledge contribution flow](https://node-template.cognidao.org/knowledge/knowledge-contribution-flow):
1. **Dated finding atom** — insert one `finding` atom for this evaluation (for example
`constraint-eval-<YYYY-MM>`) holding the date, the re-sorted scores (severity, focus
work item, evidence), the binding constraint with its 2-3 sentence explanation and
"work on now" (1-2 work items), what changed since the previous evaluation, and the
external research sources. Cite `constraint-toc-taxonomy` and the previous evaluation
atom.
2. **Taxonomy changes** — if your research surfaced missing constraint categories, propose
them in the same contribution as an update to `constraint-toc-taxonomy`, with the source
URL. Refetch the live atom first and resend every field.
3. **Never create or edit a Markdown file** for the evaluation.
### Step 7: Summarize to the user
Print:
- Binding constraint + severity + why (including external research)
- Any scores that changed and why
- Any new constraints proposed from research
- Any disagreements with previous evaluations
- What to work on now
## Principles
- **Think independently.** You are not a summarizer of internal docs. The hub atoms are inputs,
not answers. Your research and judgment are what make this skill valuable.
- **Research is mandatory.** Every evaluation must include web searches for current
landscape data. An evaluation without external research is incomplete.
- **Challenge assumptions.** If 55+ projects are all "Not Started," maybe the problem
isn't "single operator" — maybe it's "too many projects." If the team says
observability is P0, but they haven't started it in months, maybe it's not actually
their priority regardless of what they wrote down. Call out contradictions.
- **Evidence over intuition.** Every score needs evidence — but evidence can come from
external research, not just internal docs. "DAOs with voter participation below 20%
historically fail within 2 years (source)" is valid evidence for L01.
- **Identify, don't solve.** Diagnosis only. Point at the right project to work on.
Implementation is for other skills and developers.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |