specificity-constraints — Missing Constraints

Severity: INFO · Auto-fix: No · Category: 🎯 Specificity

What It Does

Detects prompts that contain a task instruction but no constraint keywords. Without constraints — "do not", "must", "limit", "exactly" — the model makes its own choices about scope, format, and edge case handling.

Trigger Conditions

Three conditions must all be true

1. The prompt is longer than 80 characters
2. The prompt contains a task verb: write, create, generate, list, explain
3. The prompt does not contain a constraint keyword: must, should, limit, maximum, minimum, between, exactly, only

This is keyword-based heuristic detection. The rule doesn't understand whether your constraints are semantically sufficient — it just checks if constraint vocabulary is present.

Examples

Triggers the rule

Write a Python function that parses configuration files and returns
a dictionary of the settings with their values.

Has write (task verb), > 80 chars, no constraint keywords.

[ INFO ] specificity-constraints (line -)
  Consider adding constraints (length, format, scope) for clearer results.

Passes the rule

Write a Python function that parses configuration files and returns
a dictionary of the settings.
Constraints:
- Must use only the standard library (no third-party packages)
- Should handle missing files gracefully (return empty dict, not raise)
- Maximum 30 lines of code
- Only support INI and JSON formats

Has must, should, maximum, only → no finding.

False Positives

Prompts with implicit constraints via examples — "Write a function. Example: validate("a@b.c") → True" has implicit format constraints from the example, but the rule still fires because no constraint keyword is present. This is intentional — explicit constraints are clearer.

should in non-constraint context — "Write a greeting that should feel warm" contains should, so the rule is suppressed even though there aren't meaningful constraints. The heuristic isn't semantically aware.

Short prompts — "Write a haiku about autumn" is < 80 chars and won't fire even though it has no constraints. The rule targets longer, more complex prompts where unconstrained generation is riskier.

Configuration

rules:
  specificity_constraints: true

# Promote to WARN:
rules:
  specificity_constraints:
    enabled: true
    level: warn

Disable:

rules:
  specificity_constraints: false

What Makes a Good Constraint

Effective constraints are negative and boundary-defining, not just positive instructions:

Type	Example
Scope limit	"Only extract from the provided text — do not use outside knowledge"
Format boundary	"Return JSON only — no prose before or after"
Length constraint	"Maximum 3 bullet points"
Library constraint	"Use only the standard library"
Error behavior	"Return null on failure — do not raise exceptions"
Exclusion	"Do not include PII in the output"

A useful mental model: what would you be annoyed to see the model do? Each of those is a constraint.

<constraints>
- Do not add docstrings or inline comments
- Do not use f-strings — use .format() for compatibility
- Do not print to stdout
- Return None (not False, not "") when the input is empty
- Maximum 25 lines
</constraints>