arifOS — Constitutional AI Kernel

Memory Governor · Kernel

arif_memory

Destructive

KERNEL 555 · Memory governor — L1–L6 governed semantic recall, storage, and persistence.

Input Schema

TableJSON Schema

Name	Required	Default
`mode`	No	recall
`tier`	No
`query`	No
`scope`	No
`top_k`	No
`aspect`	No
`hybrid`	No
`policy`	No
`cascade`	No
`content`	No
`include`	No
`payload`	No
`seal_id`	No
`to_tier`	No
`actor_id`	No
`lease_id`	No
`metadata`	No
`trace_id`	No
`_envelope`	No
`from_tier`	No
`memory_id`	No
`tier_hint`	No
`timestamp`	No
`provenance`	No
`redact_pii`	No
`session_id`	No
`structured`	No
`graph_first`	No
`new_content`	No
`truth_class`	No
`caller_chain`	No
`future_value`	No
`memory_class`	No
`policy_basis`	No
`applicability`	No
`embedding_ref`	No
`include_proof`	No
`session_token`	No
`vault_version`	No
`human_approval`	No
`new_structured`	No
`temporal_as_of`	No
`tombstone_text`	No
`idempotency_key`	No
`new_truth_class`	No
`resolution_kind`	No
`source_receipts`	No
`correction_event`	No
`memory_authority`	No
`promotion_reason`	No
`include_contested`	No
`progressive_level`	No
`require_human_ack`	No
`time_window_hours`	No
`decision_lifecycle`	No
`organ_staleness_band`	No
`supersedes_memory_id`	No
`minimised_vault_record`	No
`required_floors_satisfied`	No

Schema Changelog

Changes observed during successful MCP inspections. Dates show when Glama detected each change.

Changed4 schema fields changedJul 12, 2026

addedInput schema / properties / applicability

Added value: +{
+  "anyOf": [
+    {
+      "additionalProperties": true,
+      "type": "object"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "default": null
+}

addedInput schema / properties / decision_lifecycle

Added value: +{
+  "anyOf": [
+    {
+      "additionalProperties": true,
+      "type": "object"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "default": null
+}

addedInput schema / properties / future_value

Added value: +{
+  "anyOf": [
+    {
+      "additionalProperties": true,
+      "type": "object"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "default": null
+}

addedInput schema / properties / memory_authority

Added value: +{
+  "anyOf": [
+    {
+      "additionalProperties": true,
+      "type": "object"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "default": null
+}

Changed1 schema field changedJul 9, 2026

addedInput schema / properties / session_token

Added value: +{
+  "anyOf": [
+    {
+      "type": "string"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "default": null
+}

AddedJul 7, 2026

TDQS

C2/5.0

Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare destructiveHint=true, readOnlyHint=false, and idempotentHint=false, but the description adds no behavioral context beyond that. It does not disclose what could be destroyed (e.g., memories), whether human approval may be required, or any side effects of the various modes.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness2/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single cryptic line, which is concise in word count but severely under-specified. It reads more like a tagline than a functional description and does not earn its place by providing useful information to an agent.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness1/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

This is an extremely complex tool with 59 parameters and 8 operation modes, plus no output schema. The description provides almost no context about the tiering system, what actions are possible, when they should be used, or what outcomes to expect. It is wholly inadequate for the tool's complexity.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters1/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 0%, and the description does not compensate. With 59 parameters, it only hints at 'L1–L6' tiers but fails to explain mode, tier, query, content, memory_id, or any other parameter semantics. The mode enum values are listed in the schema but not explained in the description.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose3/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description identifies the tool as a 'Memory governor' handling 'semantic recall, storage, and persistence' across L1–L6 tiers, which conveys a general memory-management purpose. However, it does not enumerate specific operations like inspect, promote, revise, or forget that appear in the mode enum, and it does not differentiate this tool from sibling arif_* tools.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides no guidance on when to use this tool versus siblings, nor does it explain how to choose among the eight mode values in the schema. No context about the appropriate use cases for recall, inspect, attest, remember, promote, revise, forget, or audit is given.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All arifOS — Constitutional AI Kernel tools Inspect

Try in Browser

RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

C2.8/5.0

Disambiguation4/5

Each tool targets a distinct kernel function: initialization, observation, reasoning, judgment, routing, execution, memory, and sealing. The abstract descriptions could cause slight confusion, but the KERNEL labels and verb-noun distinctions make roles reasonably clear.

Naming Consistency4/5

All tools share the 'arif_' prefix and are mostly single verbs (init, observe, think, judge, route, forge, seal) but 'memory' is a noun, breaking the pattern slightly. Naming is otherwise consistent and predictable.

Tool Count5/5

Eight tools provide a well-scoped surface for a constitutional AI kernel, covering initiation, perception, reasoning, arbitration, execution, memory, and immutable sealing. The count feels appropriate for the stated purpose without redundancy or bloat.

Completeness4/5

The kernel covers the core lifecycle: init, observe, think, judge, act, and store. Minor gaps exist, such as the absence of a tool for updating or amending constitutional rules, but the primary workflows are supported.

Memory Governor · Kernel

Input Schema

Schema Changelog

TDQS

Glama MCP Gateway

TDQS

Resources