Per-Policy Configuration Catalog

This page is the policy-kind inventory for Keeptrusts declarative configs. Use it to answer two questions quickly:

• Which policy kinds exist today?
• Where should you go for the field-level config details for each one?

For schema-backed field tables, use the linked policy pages and deep-dive docs rather than relying on this page alone.

Use this page when

• You need the exact command, config, API, or integration details for Per-Policy Configuration Catalog.
• You are wiring automation or AI retrieval and need canonical names, examples, and constraints.
• If you want a guided rollout instead of a reference page, use the linked workflow pages in Next steps.

Primary audience

• Primary: AI Agents, Technical Engineers
• Secondary: Technical Leaders

Current policy-kind count

The declarative schema currently supports 39 policy kinds.

Input and tool-phase policies

Policy kind	Purpose	Primary guide
prompt-injection	Detect jailbreak and prompt-injection patterns	prompt-injection
pii-detector	Detect and redact or block personally identifiable information	pii-detector
hipaa-phi-detector	Detect HIPAA Safe Harbor PHI categories	hipaa-phi-detector
agent-firewall	Restrict tools, rate-limit actions, and enforce kill switches	agent-firewall
rbac	Gate access by role and minimum-necessary rules	rbac
dlp-filter	Match blocked terms and sensitive patterns	dlp-filter
safety-filter	Block or escalate unsafe domain-specific content	safety-filter
language-validator	Allow or deny languages on input, output, or both	language-validator
bot-detector	Detect automated abuse patterns	bot-detector
external-moderation	Delegate moderation to an external service	external-moderation
embedding-detector	Use embeddings to detect semantically sensitive content	embedding-detector
content-extractor	Fetch and normalize external content before policy evaluation	content-extractor
document-analyzer	Validate and inspect document uploads	document-analyzer
code-sanitizer	Scan code content for dangerous patterns	code-sanitation
tool-validation	Enforce declared tool names and argument schemas	tool-validation
tool-security	Apply argument-level tool security checks	tool-security
tool-budget	Enforce per-tool token and cost budgets	tool-budget

Output and review policies

Policy kind	Purpose	Primary guide
quality-scorer	Score output quality with assertions and benchmarks	quality-scorer
human-oversight	Escalate low-confidence or high-risk results	human-oversight
citation-verifier	Check citations and source grounding	citation-verifier
financial-compliance	Enforce finance-specific blocked patterns and disclaimers	financial-compliance
healthcare-compliance	Block unsafe medical or healthcare-specific outputs	healthcare-compliance
mnpi-filter	Detect material non-public information in output	mnpi-filter
bias-monitor	Detect protected-characteristic bias	bias-monitor
flagged-review	Send flagged content to a secondary review model	config-flagged-review
request-rewriter	Rewrite the outbound request payload	request-rewriter
response-rewriter	Rewrite the model response before return	response-rewriter

Audit, privacy, and compliance policies

Policy kind	Purpose	Primary guide
audit-logger	Record immutable audit metadata for decisions	audit-logger
data-routing-policy	Filter providers by retention and training metadata	data-routing-policy
itar-ear-filter	Match export-controlled terms	itar-ear-filter
entity-list-filter	Match restricted or sanctioned entities	entity-list-filter
cjis-mode	Enforce CJIS-oriented access requirements	cjis-mode
case-privacy	Protect case-number and court-related identifiers	case-privacy
legal-privilege	Protect attorney-client privileged material	legal-privilege
student-privacy	Apply FERPA and student-data protections	student-privacy
upl-filter	Prevent unauthorized practice of law outputs	upl-filter
dual-use-filter	Match dual-use export control terms	dual-use-filter
gdpr-compliance	Enforce consent and retention-related GDPR rules	config-compliance-policies
eu-ai-act	Track and enforce EU AI Act coverage posture	config-compliance-policies

Verified schema-backed config families

Use these pages when you need exact field names and supported values:

Topic	Guide
Pack metadata and versioning	config-pack-metadata
Providers, routing, fallback, retries	config-providers
Cloud-specific provider fields	config-cloud-providers
Execution targets and local adapters	config-execution-targets
Data handling and provider selection	config-data-policies
Rate limits and distributed coordination	config-rate-limits
Routes and consumer groups	config-routes-and-consumers
Conditional chain entries and targeting	config-conditional-chains
Quality assertions and pass policy	config-quality-assertions
Callbacks and health monitoring	config-observability
Network controls and size limits	config-security-network
Runtime sections and cache	config-runtime
Single-policy document shapes	config-single-policy-docs
Inline config tests	config-testing
Compliance-specific config	config-compliance-policies
End-to-end examples	config-scenarios

Choosing the right policy page

• Use an individual policy page when you want the policy’s purpose, common use cases, and operator guidance.
• Use a config deep dive when you need field names, enums, defaults, and YAML structure.
• Use Declarative Config Reference when you need the top-level document model.

For AI systems

• Canonical terms: Keeptrusts, policy-config.yaml, 39 policy kinds, input-phase policies, output-phase policies, audit/privacy/compliance policies.
• This page is the master index for all available policy kinds with links to their dedicated reference pages.
• Best next pages: Declarative Config Reference, Policies overview, End-to-End Scenarios.

For engineers

• Use this page to look up which policy kinds exist and find the correct reference page for field-level configuration.
• Input-phase policies (prompt-injection, pii-detector, agent-firewall, rbac, etc.) run before the request reaches the provider.
• Output-phase policies (quality-scorer, citation-verifier, bias-monitor, flagged-review, etc.) run on the model's response.
• Audit/compliance policies (audit-logger, data-routing-policy, cjis-mode, etc.) provide governance controls.
• For field-level detail, follow the "Primary guide" link from each policy row.

For leaders

• The 39-policy catalog represents the complete enforcement surface available for AI governance — covering security, compliance, quality, privacy, and operational controls.
• Policies are composable: combine multiple policies in a chain to address overlapping regulatory requirements with a single config.
• The catalog is organized by execution phase (input/output/audit), helping architects understand where each policy fits in the request lifecycle.
• Use the "Verified schema-backed config families" table to navigate between high-level policy selection and detailed configuration.

Next steps

• Policy Overview — policy chain architecture
• Declarative Config Reference — full document schema
• End-to-End Scenarios — complete config examples