skills.rest
maestria-co
maestria-coOFFICIAL
Explore all skills by this author →
GitHub

common-constraints

Enforces behavioral guardrails for task planning, execution, and code modification.

Updated Mar 2, 2026
Search Tags:#reliability#constraints#guardrails#self-review#process-control#code-modification
One-click install
npx skills add https://github.com/maestria-co/ai-playbook --skill common-constraints
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: common-constraints
Source: https://github.com/maestria-co/ai-playbook/tree/main/skills/common-constraints
Command: npx skills add https://github.com/maestria-co/ai-playbook --skill common-constraints

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Universal guardrails to prevent catastrophic agent failures by enforcing evidence-driven completion, limiting scope creep, and preventing unsafe modifications. These constraints help avert rationalization loops, repeated failed attempts, and drift from agreed-upon goals.

Core Features & Use Cases

  • Constraint 1: Evidence Requirement ensures that claims of completion are supported by concrete evidence such as command output, test results, or build logs.
  • Constraint 2: Failure Escalation enforces a maximum of three attempts before surfacing next steps.
  • Constraint 3: Read-First Discipline requires reading relevant files and context before any modification.
  • Constraint 4: Convention Adherence enforces project conventions and local patterns.
  • Constraint 5: Self-Review Gate prompts a quick self-review before reporting completion.
  • Constraint 6: Scope Discipline keeps changes focused on the requested task and notes out-of-scope work.

Quick Start

Ask the agent to apply all six constraints before marking a task as done.

Frequently Asked Questions about common-constraints

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent AI agent scope creep during code modification tasks?▼

Scope discipline constraints keep code modifications focused on the requested task, explicitly flagging out-of-scope work to prevent agent drift and unintended changes.

What are the best guardrails to stop agent failure loops and rationalization?▼

Universal behavioral guardrails prevent agent failure loops by enforcing a maximum of three failed attempts before escalating next steps, stopping rationalization and repeated attempts.

How do I enforce evidence-driven task completion for autonomous agents?▼

Evidence requirement constraints enforce evidence-driven completion by demanding concrete command output, test results, or build logs to support any claims of task completion.

Can I apply self-review and convention adherence constraints to software engineering agents?▼

Yes, convention adherence enforces local project patterns during execution, while a self-review gate prompts a quick self-review check before an agent reports task completion.

Why should I require read-first discipline before allowing agent code modifications?▼

Read-first discipline requires reading relevant files and context before any code modification, ensuring agents understand existing structure and preventing unsafe changes.

What limitations exist when applying universal behavioral constraints to task planning?▼

These constraints limit agents to three failure attempts before escalation and enforce strict scope discipline, which may slow down tasks requiring rapid, exploratory code modifications.

Related Skills

Explore More Software Engineering Skills→
RayzerrekRayzerrek

unity-project-scout

Inspect Unity project version, packages, asmdef layout, and folder structure.

Community
Intermediate
miyamo2miyamo2

kiro-validate-impl

Validate cross-task feature integrations across kiro tasks for spec coverage.

Community
Advanced
changoo89changoo89
19

agent-teams

Coordinate Claude Code teammates to execute, review, and plan with delegate mode and quality gates.

Community
Advanced
SergeiGolosSergeiGolos
1

lint-and-validate

Lint and validate Node.js and Python code with ESLint, Ruff, and type checkers.

Community
Intermediate