Exercise 06: Rules, Skills & Agent Configuration¶
Objective¶
Read and annotate the three configuration primitives that control AI agent behavior: rules (.mdc), skills (SKILL.md), and agent definitions (.md). Demonstrate you understand each format's frontmatter, purpose, and when it activates.
Required Reading
- Foundation README -- Glossary (terms: rule, skill, agent, frontmatter)
- Rules | Cursor Docs --
.mdcformat, frontmatter fields, precedence hierarchy - Custom Agents | Cursor Docs -- Agent
.mdformat, AGENTS.md registry - Agent Skills | Cursor Docs -- Official docs on
SKILL.mdformat, discovery, and activation - Agent Skills Guide -- Additional guide on SKILL.md format and dynamic context discovery
The exercises and validation below work in Cursor. Use the Cursor documentation links in Required Reading.
Claude Code uses an equivalent system. Rules map to CLAUDE.md project instructions. Skills use the identical .claude/skills/ directory with the same SKILL.md format. Agent definitions are handled via sequential prompting with model selection. The concepts below apply to both; only file paths differ.
Context¶
You defined these terms in Exercise 01. Now you will read the actual files, annotate their structure, and answer conceptual questions about when each activates and how they interact.
Tasks¶
Part 1: Annotate Rules¶
Read these two rule files:
- .cursor-foundation/rules/jg-planner-first.mdc (or the parent .cursor/rules/jg-planner-first.mdc)
- Any other .mdc rule in the repo (e.g. .cursor/rules/jg-issue-workflow.mdc)
For each rule, note:
- The frontmatter fields: description, alwaysApply, globs (which may be absent)
- The body sections and what each instructs the AI to do
- When this rule would be activated (based on alwaysApply and globs)
Part 2: Annotate a Skill¶
Read this skill file:
- .cursor/skills/jg-pipeline-artifact-io/SKILL.md
Note:
- The frontmatter fields: name, description
- The body sections (directory layout, reading, writing, per-agent mapping, anti-patterns)
- How the skill is discovered and activated (by description match when the agent's task is relevant)
Part 3: Annotate an Agent¶
Read this agent definition:
- .cursor-foundation/agents/jg-planner.md
Note:
- The frontmatter fields: name, model, description, readonly
- The body sections (ROLE, PRIMARY OBJECTIVE, CORE RESPONSIBILITIES, NON-GOALS)
- How AGENTS.md references this agent in the pipeline order
Part 4: Quiz Questions¶
Answer these questions in your own words (1-3 sentences each):
- What three things determine when a rule is applied? (Hint: frontmatter fields)
- How does a skill differ from a rule in terms of activation?
- What fields must an agent definition's frontmatter contain at minimum?
- What role does AGENTS.md play in the pipeline?
Output¶
Write to docs/foundation/tutorials/outputs/06-configuration.md. Use these section headings:
## Rules
(Your annotations of the two rule files)
## Skills
(Your annotation of the skill file)
## Agents
(Your annotation of the agent file)
## Quiz Answers
1. Answer to question 1 (1-3 sentences)
2. Answer to question 2 (1-3 sentences)
3. Answer to question 3 (1-3 sentences)
4. Answer to question 4 (1-3 sentences)
Validation
python3 docs/foundation/tutorials/verify.py --exercise 06
Checks: file exists, all 4 sections present, each section mentions the correct frontmatter fields, quiz answers have sufficient depth.
Answer
Quiz Answers:
-
What three things determine when a rule is applied?
alwaysApply(if true, always injected),globs(file path patterns that trigger it), anddescription(used for relevance matching when alwaysApply is false and no globs match). -
How does a skill differ from a rule in terms of activation? Rules are passively injected (pushed by the system). Skills are actively discovered -- agents scan SKILL.md descriptions and pull in a skill only when it matches the current task.
-
What fields must an agent definition's frontmatter contain at minimum?
name(dispatch identifier),model(which AI model), anddescription(for discovery).readonlyis important for safety but may be optional. -
What role does AGENTS.md play? It is the pipeline registry: an index of all agents with models, roles, I/O, pipeline execution order, and subagent_type mappings.
See docs/foundation/tutorials/solutions/06-configuration-guide.md in the source repo for complete annotation examples.