The Mandatory Skill Check Protocol

Relevant source files

• .cursor-plugin/plugin.json
• README.md
• skills/brainstorming/SKILL.md
• skills/using-superpowers/SKILL.md

Purpose and Scope

This document describes the core enforcement mechanism that makes Superpowers work: the Mandatory Skill Check Protocol as defined in skills/using-superpowers/SKILL.md. This meta-skill is injected into every agent session and requires agents to check for applicable skills before any response or action.

For information about what skills are and how they are structured, see page 3.1 For information about how agents discover and invoke skills using platform tools, see page 3.3 For the technical details of how this rule is injected at session start, see page 4.4

---

The using-superpowers Meta-Skill

The using-superpowers skill is not a workflow skill — it is the meta-skill that governs how all other skills are used. Its content is loaded into agent context at session start and defines:

• When to invoke skills (before every response)
• How to invoke them (via the Skill tool)
• What to do after invocation (announce, create TodoWrite checklist items, follow exactly)
• How to resist skipping them (Red Flags table)

The central rule, stated in skills/using-superpowers/SKILL.md:

Invoke relevant or requested skills BEFORE any response or action.

This applies to every agent turn, including clarifying questions, codebase exploration, and implementation. The skill states this in absolute terms:

IF A SKILL APPLIES TO YOUR TASK, YOU DO NOT HAVE A CHOICE. YOU MUST USE IT.
This is not negotiable. This is not optional. You cannot rationalize your way out of this.

Sources: skills/using-superpowers/SKILL.md6-12 skills/using-superpowers/SKILL.md22-24

---

The 1% Rule

The 1% rule is the decision threshold for skill invocation. If the agent judges even a 1% chance that a skill might apply, it must invoke the skill to check.

This deliberately low threshold serves specific purposes:

Purpose	Rationale
Prevents false negatives	Agents frequently skip skills thinking "this is simple"
Counters confidence bias	High-confidence agents are most likely to skip skills that would help
Eliminates the decision	A clear threshold removes the "should I check?" judgment call
Safe false positives	Invoking an inapplicable skill is low-cost; missing a relevant skill is expensive

The threshold is stated as a probability ("even 1% chance") rather than a certainty judgment to block agent rationalization.

Sources: skills/using-superpowers/SKILL.md7-8 skills/using-superpowers/SKILL.md24

---

The Mandatory Protocol Flow

Diagram: Mandatory Skill Check Protocol — Decision Flow

The following diagram reflects the logic embedded in skills/using-superpowers/SKILL.md. It shows two entry points: a normal user message, and a special case when the agent is about to enter its native plan mode.

Sources: skills/using-superpowers/SKILL.md26-54

Protocol Steps

The mandatory steps in order:

1. Check plan mode intercept — If about to enter EnterPlanMode, verify brainstorming skill has already run
2. Evaluate skill applicability — If ≥1% chance any skill applies, proceed to invoke
3. Invoke Skill tool — Load skill content using superpowers:skill-name namespace
4. Announce usage — State "Using [skill] to [purpose]" before proceeding
5. Create TodoWrite checklist — If the skill contains a checklist, create one TodoWrite item per entry
6. Follow skill exactly — Execute the loaded skill's workflow
7. Respond — Provide output to the user only after completing the above

Sources: skills/using-superpowers/SKILL.md26-54

Plan Mode Interception

The EnterPlanMode intercept prevents a specific failure mode: agents skipping the brainstorming skill and jumping directly into Claude's native plan mode without completing the design phase. When the agent is about to enter plan mode, the protocol checks whether brainstorming has already occurred. If not, it routes through skills/brainstorming/SKILL.md first.

Sources: skills/using-superpowers/SKILL.md29-42

---

Bootstrap Mechanism: Injecting the Protocol

How the Protocol Enters Agent Context

The using-superpowers meta-skill is injected into every agent session before the agent's first response. The injection path differs by platform:

Diagram: Bootstrap Injection by Platform — Code Entities

Sources: skills/using-superpowers/SKILL.md1-96 .claude-plugin/hooks.json hooks/session-start.sh

Synchronous Hook Execution

For Claude Code, the hook is executed synchronously ("async": false in hooks.json). This guarantees the bootstrap content is in context before the agent's first turn, preventing the failure mode where an agent responds before the Rule is loaded.

Sources: .claude-plugin/hooks.json

Bootstrap Content Structure

The content loaded from skills/using-superpowers/SKILL.md includes these components:

Component	Lines	Purpose
EXTREMELY-IMPORTANT block	6-12	Absolute language enforcing mandatory invocation
## How to Access Skills	14-18	Tells agent to use Skill tool, never Read
## The Rule	22-24	Core one-sentence protocol statement
Skill flow diagram (DOT)	26-53	Visual protocol flowchart
## Red Flags table	56-73	Pre-empts common rationalizations
## Skill Priority	75-82	Ordering when multiple skills apply
## Skill Types	84-91	Rigid vs. flexible skill handling
## User Instructions	93-95	Instructions say WHAT, not HOW

Sources: skills/using-superpowers/SKILL.md1-96

---

Red Flags: Rationalization Prevention

The Rationalization Problem

A core challenge in skill enforcement is agent rationalization — agents generating plausible-sounding reasons to skip skills. Common examples: "this is simple," "I remember this skill," or "let me gather context first."

The ## Red Flags section in skills/using-superpowers/SKILL.md pre-emptively counters these patterns by placing rebuttals directly in agent context.

Red Flags Reference Table

Defined at skills/using-superpowers/SKILL.md58-73:

Red Flag Thought	Reality
"This is just a simple question"	Questions are tasks. Check for skills.
"I need more context first"	Skill check comes BEFORE clarifying questions.
"Let me explore the codebase first"	Skills tell you HOW to explore. Check first.
"I can check git/files quickly"	Files lack conversation context. Check for skills.
"Let me gather information first"	Skills tell you HOW to gather information.
"This doesn't need a formal skill"	If a skill exists, use it.
"I remember this skill"	Skills evolve. Read current version.
"This doesn't count as a task"	Action = task. Check for skills.
"The skill is overkill"	Simple things become complex. Use it.
"I'll just do this one thing first"	Check BEFORE doing anything.
"This feels productive"	Undisciplined action wastes time. Skills prevent this.
"I know what that means"	Knowing the concept ≠ using the skill. Invoke it.

Sources: skills/using-superpowers/SKILL.md56-73

---

TodoWrite Checklist Integration

When an invoked skill contains a numbered checklist, the protocol requires the agent to convert each checklist item into a TodoWrite task before beginning execution. This keeps progress visible and prevents the agent from skipping steps mid-skill.

Example: skills/brainstorming/SKILL.md has a 6-item checklist:

1. Explore project context
2. Ask clarifying questions
3. Propose 2-3 approaches
4. Present design (per section)
5. Write design doc
6. Transition to implementation

When the agent invokes brainstorming, the protocol requires six corresponding TodoWrite entries to be created before any step is executed.

Diagram: TodoWrite Integration — Skill Checklist to Task Tracking

Sources: skills/using-superpowers/SKILL.md48-53 skills/brainstorming/SKILL.md22-31

Skill Priority System

When Multiple Skills Apply

When multiple skills could apply, the protocol at skills/using-superpowers/SKILL.md75-82 defines a fixed priority order:

1. Process skills first (brainstorming, systematic-debugging) — determine HOW to approach the task
2. Implementation skills second (frontend-design, mcp-builder) — guide execution details

Priority Decision Examples

User Request	First Skill	Reason
"Let's build X"	brainstorming	Design before implementation
"Fix this bug"	systematic-debugging	Investigation before fixes
"Add feature Y"	brainstorming	Requirements before code

Sources: skills/using-superpowers/SKILL.md75-82

---

Code Integration Points

Skill Tool Integration

Agents invoke skills using the platform's native Skill tool. The bootstrap content in skills/using-superpowers/SKILL.md14-18 explicitly instructs:

In Claude Code: Use the Skill tool. When you invoke a skill, its content is loaded and presented to you — follow it directly. Never use the Read tool on skill files.

This prevents agents from reading the skill file directly via Read instead of going through the Skill tool.

Slash Commands

Three slash commands in commands/ redirect user-facing shortcuts to skills:

Command	Skill Invoked
/brainstorm	superpowers:brainstorming
/write-plan	superpowers:writing-plans
/execute-plan	superpowers:executing-plans

All three have disable-model-invocation: true, meaning only users (not agents) can trigger them.

Sources: skills/using-superpowers/SKILL.md14-18 commands/

---

Summary

The Rule and Mandatory First Response Protocol form the enforcement backbone of Superpowers:

1. The Rule requires skill invocation before any response
2. 1% threshold sets a deliberately low bar for "might apply"
3. Bootstrap mechanism injects the protocol at session start (synchronous)
4. Red Flags table pre-emptively counters common rationalizations
5. Skill priority guides agents when multiple skills apply
6. Plan mode intercept prevents native planning without brainstorming

This multi-layered approach treats skill compliance as a technical constraint enforced through context injection, explicit protocols, and anticipatory rationalization prevention.