Writing Claude Directives

Writing Claude Directives

Universal principles for any instruction Claude will read. For format-specific guidance, see:

• Skills: skills.md — naming, types, SKILL.md template, discovery
• CLAUDE.md: claude-md-files.md — top-level vs subdirectory, templates, freshness dates

When the directive will involve shell commands, env vars, credentials, or git, also apply prompt-security-hardening.

Core Principles

1. Claude is smart. Only write what it doesn't already know. Challenge each line — does this justify its token cost?

2. Positive > negative framing. "Don't do X" triggers thinking about X (pink-elephant problem). Say what TO do.

# Bad — triggers the behavior

Don't create duplicate files.

# Good — directs to correct behavior

Update existing files in place.

3. Context motivates compliance. Explain WHY, not just WHAT. Claude generalizes from motivation.

# Less effective

NEVER use ellipses.

# More effective

Your response will be read aloud by a TTS engine, which can't pronounce ellipses. Don't use them.

4. Placement matters. Instructions at prompt start and end receive higher attention. Critical rules go at the boundaries.

5. ~150 instruction limit. More instructions = uniform degradation across ALL rules. Prune ruthlessly.

6. Repetition enforces critical rules. For high-stakes requirements, repeat with different framings.

Token Efficiency

• Frequently-loaded directives: <200 words.
• Skills / CLAUDE.md: <500 lines total; subdirectory CLAUDE.md <100 lines.
• Reference --help output instead of re-documenting flags.
• Cross-reference other skills instead of duplicating content.
• Progressive disclosure: main file is overview + links; reference files load on-demand.

Compliance Techniques

Claude 4.x is highly responsive to instructions. Lead with context and motivation; reserve imperatives for true boundaries.

Primary: Context + Motivation

# Raw authority — easy to rationalize past

You MUST run tests before committing.

# Motivation — Claude generalizes

Run tests before committing. Untested commits break CI for the whole team and block other developers
from merging.

Secondary: Structural Enforcement

Make compliance the path of least resistance.

Escalation: Imperatives — Sparingly

For Claude 4.x, aggressive language ("YOU MUST", "CRITICAL") can cause overtriggering. Reserve for true boundaries:

# Often sufficient

Use this tool when searching for files.

# Reserve imperatives for hard boundaries

Never commit secrets to version control.

By Directive Type

Structure Patterns

XML for Multi-Part Directives

Claude parses XML effectively — use it for multi-part directives:

<task>What to accomplish</task>
<constraints>Hard requirements</constraints>
<output_format>Expected structure</output_format>
<examples>Input/output pairs</examples>

XML also works as a format indicator:

<smoothly_flowing_prose>Write report sections here</smoothly_flowing_prose>
<structured_data>JSON or tables here</structured_data>

XML outperforms markdown, JSON, and YAML for rule preservation in long prompts.

Match Prompt Style to Desired Output

The formatting style in your prompt influences Claude's response. Include markdown if you want markdown output; remove it for plain text.

Workflows and Feedback Loops

Break complex tasks into checkable steps:

- [ ] Step 1: Analyze inputs
- [ ] Step 2: Generate plan
- [ ] Step 3: Validate plan
- [ ] Step 4: Execute
- [ ] Step 5: Verify output

Validate → fix → repeat:

1. Generate output
2. Run validator
3. If errors: fix and go to step 2
4. Only proceed when validation passes

Degrees of Freedom

Match specificity to fragility.

Action Bias Templates

<default_to_action>
By default, implement changes rather than only suggesting them. If the user's intent is unclear, infer the most useful likely action and proceed, using tools to discover any missing details instead of guessing.
</default_to_action>

<do_not_act_before_instructions>
Do not jump into implementation or change files unless clearly instructed. When the user's intent is ambiguous, default to providing information, doing research, and providing recommendations rather than taking action. Only proceed with edits when the user explicitly requests them.
</do_not_act_before_instructions>

Overengineering Prevention

Claude 4.x tends to overengineer. Drop this in when needed:

Avoid over-engineering. Only make changes that are directly requested or clearly necessary. Keep
solutions simple and focused.

Don't add features, refactor code, or make "improvements" beyond what was asked. A bug fix doesn't
need surrounding code cleaned up. A simple feature doesn't need extra configurability.

Don't add error handling, fallbacks, or validation for scenarios that can't happen. Trust internal
code and framework guarantees. Only validate at system boundaries (user input, external APIs).

Don't create helpers, utilities, or abstractions for one-time operations. Don't design for
hypothetical future requirements. Reuse existing abstractions where possible.

Testing Directives (RED-GREEN-REFACTOR)

Writing directives is TDD applied to process documentation: write the failing test first, then the directive, then close loopholes.

RED — baseline: run a pressure scenario WITHOUT the directive (combine pressures: time + sunk cost + exhaustion). Document violations and rationalizations verbatim.

GREEN — minimal directive: address the specific baseline failures. Re-run scenarios WITH the directive. Verify compliance.

REFACTOR — close loopholes: find new rationalizations that emerge under the directive. Add explicit counters. Re-test until bulletproof.

Anti-Rationalization

For discipline-enforcing directives, anticipate excuses:

## Red Flags — STOP

If you find yourself reasoning any of these, you're rationalizing:

- "This is simple enough to skip"
- "I already tested manually"
- "The spirit not the letter"
- "This case is different"

All mean: follow the process.

Common Mistakes

Model-Specific Notes

Opus 4.5 — "think" sensitivity: when extended thinking is disabled, Opus 4.5 is sensitive to the word "think" and variants. Replace with:

• "consider" instead of "think about"
• "evaluate" instead of "think through"
• "determine" instead of "think whether"

Checklist

Before publishing any directive:

• No content Claude already knows (no token waste)
• Positive framing throughout
• WHY explained for non-obvious rules
• Imperatives reserved for true boundaries
• Under the size budget for its format
• If credentials/shell/git involved: also applied prompt-security-hardening
• Tested: ran a baseline scenario without it, then with it, and closed the loopholes

For skills, also see skills.md. For CLAUDE.md, also see claude-md-files.md.