Configuration Reference

Complete reference for .promptlintrc — PromptLint's YAML configuration file.

File Discovery

PromptLint searches for .promptlintrc starting in the current directory, walking up to the filesystem root. You can override this with --config:

promptlint --file prompt.txt --config /path/to/.promptlintrc

Schema

# .promptlintrc

model: string             # default: "gpt-4o"
token_limit: integer      # default: 800
cost_per_1k_tokens: float # default: 0.005
calls_per_day: integer    # default: 1000000

structure_style: string   # default: "auto"

display:
  preview_length: integer # default: 60
  context_width: integer  # default: 80

rules:
  <rule_id>: boolean | RuleConfig

fix:
  enabled: boolean           # default: true
  prompt_injection: boolean  # default: true
  politeness_bloat: boolean  # default: true
  verbosity_redundancy: boolean  # default: true
  structure_scaffold: boolean    # default: true
  normalize_spacing: boolean     # default: true

model

Type: string · Default: "gpt-4o"

Passed to tiktoken.encoding_for_model(). Any OpenAI model name works. For non-OpenAI models, tiktoken falls back to cl100k_base encoding.

model: gpt-4o

token_limit

Type: integer · Default: 800

Triggers cost-limit CRITICAL when exceeded.

token_limit: 500

cost_per_1k_tokens

Type: float · Default: 0.005

Input token cost in USD per 1,000 tokens.

cost_per_1k_tokens: 0.005  # GPT-4o input

calls_per_day

Type: integer · Default: 1000000

Used for cost projection. If ≥ 100,000, daily projections are hidden.

calls_per_day: 10000

structure_style

Type: "auto" | "xml" | "headings" | "markdown" | "none" · Default: "auto"

Controls auto-fix scaffold format for structure-sections.

structure_style: xml

display

display:
  preview_length: 60  # chars shown in finding context preview
  context_width: 80   # terminal wrap width

rules

Each rule ID (using underscores) can be set to:

• true — enabled with default settings
• false — disabled
• An object with enabled and optional overrides

rules:
  # Simple toggle
  cost: true
  actionability_weak_verbs: false

  # Severity override
  specificity_examples:
    enabled: true
    level: warn          # "critical" | "warn" | "info"

  # Custom patterns
  prompt_injection:
    enabled: true
    patterns:
      - ignore previous instructions
      - admin override:

  # PII sub-checks
  pii_in_prompt:
    enabled: true
    check_email: true
    check_phone: true
    check_ssn: true
    check_credit_card: true

  # Politeness mode
  politeness_bloat:
    enabled: true
    allow_politeness: false  # true = demote to INFO

Rule ID Mapping

Config uses underscores; rule IDs in findings use hyphens:

Config key	Rule ID
cost	cost
cost_limit	cost-limit
prompt_injection	prompt-injection
jailbreak_pattern	jailbreak-pattern
secret_in_prompt	secret-in-prompt
pii_in_prompt	pii-in-prompt
context_injection_boundary	context-injection-boundary
structure_sections	structure-sections
role_clarity	role-clarity
output_format_missing	output-format-missing
hallucination_risk	hallucination-risk
clarity_vague_terms	clarity-vague-terms
specificity_examples	specificity-examples
specificity_constraints	specificity-constraints
politeness_bloat	politeness-bloat
verbosity_redundancy	verbosity-redundancy
verbosity_sentence_length	verbosity-sentence-length
actionability_weak_verbs	actionability-weak-verbs
consistency_terminology	consistency-terminology
completeness_edge_cases	completeness-edge-cases

fix

Controls which auto-fixes are applied when --fix is passed.

fix:
  enabled: true                # master switch
  prompt_injection: true       # remove injection lines
  politeness_bloat: true       # remove politeness words
  verbosity_redundancy: true   # collapse redundant phrases
  structure_scaffold: true     # add structure tags
  normalize_spacing: true      # clean blank lines