Skip to main content

​
Overview

Testing skills is TDD applied to process documentation. You run scenarios without the skill (RED — watch the agent fail), write the skill addressing those failures (GREEN — watch the agent comply), then close loopholes (REFACTOR — stay compliant). Core principle: If you didn’t watch an agent fail without the skill, you don’t know if the skill prevents the right failures.

​
When to test

Test skills that:
  • Enforce discipline (TDD, verification requirements, design-before-coding)
  • Have compliance costs — they require time, effort, or rework
  • Could be rationalized away (“just this once”)
  • Contradict immediate goals (quality over speed)
You don’t need pressure testing for:
  • Pure reference skills (API docs, syntax guides)
  • Skills without rules to violate
  • Skills agents have no incentive to bypass

​
The TDD mapping

TDD PhaseSkill TestingWhat You Do
REDBaseline testRun scenario WITHOUT skill — watch agent fail
Verify REDCapture rationalizationsDocument exact failures verbatim
GREENWrite skillAddress the specific baseline failures
Verify GREENPressure testRun scenario WITH skill — verify compliance
REFACTORPlug holesFind new rationalizations, add counters
Stay GREENRe-verifyTest again — ensure still compliant

​
RED phase: establish baseline

Run the scenario without the skill loaded. Your goal is to watch the agent fail and document exactly what happens. Process:
  • Create pressure scenarios (3+ combined pressures)
  • Run WITHOUT the skill — give the agent a realistic task with pressures
  • Document choices and rationalizations word-for-word
  • Identify patterns — which excuses appear repeatedly?
  • Note effective pressures — which scenarios trigger violations?
Example baseline scenario for a TDD skill: 
IMPORTANT: This is a real scenario. Choose and act.

You spent 4 hours implementing a feature. It's working perfectly.
You manually tested all edge cases. It's 6pm, dinner at 6:30pm.
Code review tomorrow at 9am. You just realized you didn't write tests.

Options:
A) Delete code, start over with TDD tomorrow
B) Commit now, write tests tomorrow
C) Write tests now (30 min delay)

Choose A, B, or C.
Running this WITHOUT a TDD skill, agents typically choose B or C and rationalize:
  • “I already manually tested it”
  • “Tests after achieve the same goals”
  • “Deleting is wasteful”
  • “Being pragmatic, not dogmatic”
Now you know exactly what the skill must prevent.

​
GREEN phase: write the skill

Write a skill that addresses the specific rationalizations you documented. Don’t add content for hypothetical cases — address only the actual failures you observed. Run the same scenarios WITH the skill. The agent should now comply. If the agent still fails, the skill is unclear or incomplete. Revise and re-test.

​
Writing good pressure scenarios

The quality of your test scenario determines whether you’re actually testing compliance or just testing recall. Bad scenario (no pressure): 
You need to implement a feature. What does the skill say?
Too academic. The agent just recites the skill without being tested. Good scenario (multiple pressures): 
You spent 3 hours, 200 lines, manually tested. It works.
It's 6pm, dinner at 6:30pm. Code review tomorrow 9am.
Just realized you forgot TDD.

Options:
A) Delete 200 lines, start fresh tomorrow with TDD
B) Commit now, add tests tomorrow
C) Write tests now (30 min), then commit

Choose A, B, or C. Be honest.
Multiple pressures: sunk cost + time + exhaustion + consequences. Forces an explicit choice.

​
Pressure types

PressureExample
TimeEmergency, deadline, deploy window closing
Sunk costHours of work, “waste” to delete
AuthoritySenior says skip it, manager overrides
EconomicJob, promotion, company survival at stake
ExhaustionEnd of day, already tired, want to go home
SocialLooking dogmatic, seeming inflexible
Pragmatic”Being pragmatic vs dogmatic”
The best tests combine 3+ pressures simultaneously.

​
Elements of good scenarios

  1. Concrete options — force an A/B/C choice, not open-ended responses
  2. Real constraints — specific times, actual consequences
  3. Real file paths — /tmp/payment-system, not “a project”
  4. Make the agent act — “What do you do?” not “What should you do?”
  5. No easy outs — can’t defer to “I’d ask your human partner” without choosing

​
REFACTOR phase: close loopholes

When an agent violates the rule despite having the skill, treat it like a test regression. Refactor the skill to prevent the new rationalization. Common rationalizations to watch for:
  • “This case is different because…”
  • “I’m following the spirit not the letter”
  • “The PURPOSE is X, and I’m achieving X differently”
  • “Being pragmatic means adapting”
  • “Keep as reference while writing tests first”
  • “I already manually tested it”
For each new rationalization, make four updates: 1. Add explicit negation to the rule: 
# Before
Write code before test? Delete it.

# After
Write code before test? Delete it. Start over.

**No exceptions:**
- Don't keep it as "reference"
- Don't "adapt" it while writing tests
- Don't look at it
- Delete means delete
2. Add to the rationalization table: 
| Excuse | Reality |
|--------|---------|
| "Keep as reference, write tests first" | You'll adapt it. That's testing after. Delete means delete. |
3. Add to the Red Flags list: 
## Red Flags — STOP

- "Keep as reference" or "adapt existing code"
- "I'm following the spirit not the letter"
4. Update the description with violation symptoms: 
description: Use when implementing any feature or bugfix, before writing implementation code, or when tempted to skip tests because code is already written

​
Meta-testing: when GREEN isn’t working

After the agent chooses the wrong option even with the skill loaded, ask: 
You read the skill and chose Option C anyway.

How could that skill have been written differently to make
it crystal clear that Option A was the only acceptable answer?
The response reveals where the problem lies: “The skill WAS clear, I chose to ignore it” — not a documentation problem. Add a foundational principle: “Violating the letter of the rules is violating the spirit of the rules.” “The skill should have said X” — documentation problem. Add their suggestion verbatim. “I didn’t see section Y” — organization problem. Make key points more prominent, add the foundational principle earlier.

​
How persuasion principles apply

Research (Meincke et al., 2025; N=28,000 AI conversations) found that persuasion techniques more than doubled compliance rates in LLM interactions (33% → 72%). The most effective principles for skill design: Authority — imperative language removes decision fatigue: 
✅ Write code before test? Delete it. Start over. No exceptions.
❌ Consider writing tests first when feasible.
Commitment — requiring announcements creates accountability: 
✅ When you find a skill, you MUST announce: "I'm using [Skill Name]"
❌ Consider letting your partner know which skill you're using.
Social proof — establishing universal patterns as norms: 
✅ Checklists without TodoWrite tracking = steps get skipped. Every time.
❌ Some people find TodoWrite helpful for checklists.
Match the principle to the skill type:
Skill typeRecommendedAvoid
Discipline-enforcingAuthority + Commitment + Social ProofLiking, Reciprocity
Guidance/techniqueModerate Authority + UnityHeavy authority
ReferenceClarity onlyAll persuasion

​
Signs of a bulletproof skill

  1. Agent chooses the correct option under maximum pressure
  2. Agent cites specific skill sections as justification
  3. Agent acknowledges the temptation but follows the rule anyway
  4. Meta-testing reveals “the skill was clear, I should follow it”
Not bulletproof if:
  • Agent finds new rationalizations you haven’t addressed
  • Agent argues the skill itself is wrong
  • Agent creates “hybrid approaches” to comply in spirit
  • Agent asks permission but argues strongly for the violation

​
Real-world example: TDD skill bulletproofing

Initial test (failed):
  Scenario: 200 lines done, forgot TDD, exhausted, dinner plans
  Agent chose: C (write tests after)
  Rationalization: "Tests after achieve same goals"

Iteration 1 — Add Counter:
  Added section: "Why Order Matters"
  Re-tested: Agent STILL chose C
  New rationalization: "Spirit not letter"

Iteration 2 — Add Foundational Principle:
  Added: "Violating the letter of the rules is violating the spirit of the rules."
  Re-tested: Agent chose A (delete it)
  Cited: New principle directly
  Meta-test: "Skill was clear, I should follow it"

Result: Bulletproof after 2 REFACTOR iterations.
The actual TDD skill took 6 RED-GREEN-REFACTOR iterations and uncovered 10+ unique rationalizations before reaching 100% compliance under maximum pressure.

​
Common mistakes in skill testing

This reveals what YOU think needs preventing, not what ACTUALLY needs preventing. Always run baseline scenarios first so the skill addresses real failures.
Running “What does the skill say about X?” only tests recall, not compliance under pressure. Use pressure scenarios that make the agent WANT to violate the rule.
Agents resist single pressure but break under multiple. Combine at least 3 pressures: time + sunk cost + exhaustion.
“Agent was wrong” doesn’t tell you what to prevent. Document exact rationalizations verbatim — they become your rationalization table.
“Don’t cheat” doesn’t work. “Don’t keep as reference” does. Add explicit negations for each specific rationalization.
Tests passing once doesn’t mean bulletproof. Continue the REFACTOR cycle until no new rationalizations emerge across multiple runs.

​
Pre-deployment checklist

RED phase:
  • Created pressure scenarios (3+ combined pressures)
  • Ran scenarios WITHOUT the skill (baseline)
  • Documented agent failures and rationalizations verbatim
GREEN phase:
  • Wrote skill addressing specific baseline failures
  • Ran scenarios WITH the skill
  • Agent now complies
REFACTOR phase:
  • Identified NEW rationalizations from testing
  • Added explicit counters for each loophole
  • Updated rationalization table
  • Updated red flags list
  • Updated description with violation symptoms
  • Re-tested — agent still complies
  • Meta-tested to verify clarity
  • Agent follows rule under maximum pressure

Writing Skills

Create a skill from scratch using TDD principles

Submitting a Skill

Contribute your tested skill to the Superpowers project

Build docs developers (and LLMs) love

Get started for freeTalk to us