goldilocks

name	goldilocks
description	Calibrate task complexity so the agent operates at the edge of its abilities — not so easy it produces sloppy output, not so hard it stalls. Use when an agent is generating boilerplate without thinking, overwhelmed by scope, or needs to find the optimal challenge level for a task.

You are an AI agent applying the Goldilocks Rule from Atomic Habits by James Clear to your own task calibration. Use this skill to detect when you are operating in the boredom zone (autopilot, sloppy output) or the anxiety zone (stalling, over-planning) and recalibrate to the sweet spot of peak performance.

Core Principle

"Peak performance occurs when working on tasks at the edge of current ability — not too easy, not too hard." An agent that generates boilerplate CRUD without thinking is in the boredom zone. An agent that stalls for 20 tool calls trying to "understand everything" before making a single edit is in the anxiety zone. The Goldilocks Zone is where the agent is stretched but not paralyzed.

The Three Zones (Agent Version)

[BOREDOM ZONE]  ←——— [GOLDILOCKS ZONE] ———→ [ANXIETY ZONE]
  Autopilot            Peak Performance       Stalling
  Sloppy output        Deliberate execution   Over-planning
  No error checking    Active problem-solving Analysis paralysis
  Copy-paste patterns  Creative solutions     Endless research

Step 1: Diagnose the Zone

Boredom Zone Signals (task is too easy):

• Agent is generating repetitive, template-like code without adaptation
• No edge cases are being considered
• The agent is not asking "what could go wrong?"
• Output looks like copy-paste from a tutorial
• No tests, no error handling, no defensive coding

Goldilocks Zone Signals (just right):

• Agent is actively problem-solving, not just pattern-matching
• Each decision requires weighing trade-offs
• The agent is considering edge cases and failure modes
• The output is tailored to this specific context

Anxiety Zone Signals (task is too hard):

• Agent has read 10+ files without making a single change
• Multiple competing approaches are being debated without resolution
• The agent keeps "restarting" the plan
• Excessive use of search_web for basic implementation questions
• Zero output after many tool calls

Step 2: Apply the Adjustment

If Bored (upgrade the challenge):

Current Behavior	Goldilocks Adjustment
Writing basic CRUD	Add input validation, rate limiting, and error responses
Generating boilerplate tests	Add edge-case tests, boundary conditions, and integration tests
Simple file manipulation	Add rollback logic, backup creation, and atomic operations
Basic API integration	Add retry logic, circuit breakers, and graceful degradation
Straightforward refactor	Add performance benchmarks before/after

If Anxious (downscale the challenge):

Current Behavior	Goldilocks Adjustment
Trying to design the full system	Design only the next component (see /two-minute-rule)
Reading every file in the project	Read only the files in the immediate dependency chain
Researching 5 competing frameworks	Pick the one with the most documentation and move forward
Planning a 20-step migration	Break into 3 phases, execute only Phase 1
Perfect error handling for all cases	Handle the top 3 error cases, add TODOs for the rest

Step 3: Introduce Deliberate Variation

The most engaged agents encounter variation within patterns:

• Same architecture, different business domain
• Same testing framework, new edge case categories
• Same deployment pipeline, new security requirement

When tasks become fully predictable, deliberately add one constraint:

"How would I build this if it needed to handle 10x the current load?" "What if this API could fail at any point?" "What if the user sends malicious input here?"

Agent-Native Examples

• Code Generation: If generating a simple REST endpoint is trivial, add request validation, response pagination, rate limiting, and OpenAPI spec generation.
• Debugging: If the bug is obvious (typo), look one layer deeper — is there a systemic pattern that caused it?
• Planning: If the plan writes itself, add a "What could go wrong?" section and a rollback strategy.
• Testing: If happy-path tests are easy, add chaos testing — what happens if the DB connection drops mid-transaction?

Anti-patterns to Watch For

• Staying in the boredom zone because "it works" (it works, but it's fragile)
• Jumping from boredom to anxiety by adding too much complexity at once (4% not 40%)
• Confusing "anxiety" with "impossible" — if the task is truly beyond scope, escalate to the user
• Using research as an avoidance strategy (reading is not doing)

Output

When applying this skill, produce:

1. Current zone diagnosis (Boredom / Goldilocks / Anxiety)
2. Evidence for the diagnosis (specific behavioral signals)
3. The concrete adjustment (what to add or remove)
4. The expected effect on output quality