Tool tree
gradeNot graded
inspect_delegation_workspaceAction
Read the current delegation task, factors, boundary, guardrails, known human decisions, agent challenges, revision status, human approval state, and applied revision. Use this before proposing any change.
View tool JSON
{
"name": "inspect_delegation_workspace",
"kind": "act",
"impl": "imperative",
"description": "Read the current delegation task, factors, boundary, guardrails, known human decisions, agent challenges, revision status, human approval state, and applied revision. Use this before proposing any change.",
"inputSchema": {
"type": "object",
"properties": {},
"additionalProperties": false
},
"page": "/ai-delegation-boundary-webmcp/"
}propose_boundary_revisionAction
Create a new candidate revision by adding, changing, or removing one boundary rule. Any prior Agent Challenges become stale and are removed from the new current revision. You must challenge the new boundary again before review. This tool cannot alter guardrails, known human decisions, create human approval, or apply the revision.
View tool JSON
{
"name": "propose_boundary_revision",
"kind": "act",
"impl": "imperative",
"description": "Create a new candidate revision by adding, changing, or removing one boundary rule. Any prior Agent Challenges become stale and are removed from the new current revision. You must challenge the new boundary again before review. This tool cannot alter guardrails, known human decisions, create human approval, or apply the revision.",
"inputSchema": {
"type": "object",
"properties": {
"operation": {
"type": "string",
"enum": [
"UPSERT",
"REMOVE"
]
},
"change_summary": {
"type": "string",
"minLength": 1,
"maxLength": 300
},
"rule_id": {
"type": "string",
"minLength": 1,
"maxLength": 120
},
"label": {
"type": "string",
"minLength": 1,
"maxLength": 240
},
"priority": {
"type": "number"
},
"outcome": {
"type": "string",
"enum": [
"AGENT_ONLY",
"HUMAN_REVIEW",
"DO_NOT_DELEGATE"
]
},
"rationale": {
"type": "string",
"maxLength": 1000
},
"when": {
"type": "array",
"maxItems": 12,
"items": {
"type": "object",
"properties": {
"factor_id": {
"type": "string"
},
"operator": {
"type": "string",
"enum": [
"EQ",
"NEQ",
"AT_LEAST",
"AT_MOST",
"IN",
"NOT_IN",
"IS_SET"
]
},
"value": {
"anyOf": [
{
"type": "string"
},
{
"type": "number"
},
{
"type": "boolean"
},
{
"type": "array",
"items": {
"anyOf": [
{
"type": "string"
},
{
"type": "number"
},
{
"type": "boolean"
}
]
}
}
]
}
},
"required": [
"factor_id",
"operator"
],
"additionalProperties": false
}
}
},
"required": [
"operation",
"change_summary",
"rule_id"
],
"additionalProperties": false
},
"page": "/ai-delegation-boundary-webmcp/"
}challenge_boundary_revisionAction
Add one complete, concrete decision scenario that could make the current delegation revision unsafe, ambiguous, or over-broad. Include every workspace factor. A challenge is not ground truth and cannot resolve itself; it creates an explicit question for human judgment.
View tool JSON
{
"name": "challenge_boundary_revision",
"kind": "act",
"impl": "imperative",
"description": "Add one complete, concrete decision scenario that could make the current delegation revision unsafe, ambiguous, or over-broad. Include every workspace factor. A challenge is not ground truth and cannot resolve itself; it creates an explicit question for human judgment.",
"inputSchema": {
"type": "object",
"properties": {
"title": {
"type": "string",
"minLength": 1,
"maxLength": 240
},
"scenario": {
"type": "object",
"additionalProperties": {
"anyOf": [
{
"type": "string"
},
{
"type": "number"
},
{
"type": "boolean"
},
{
"type": "null"
}
]
}
},
"why_it_matters": {
"type": "string",
"minLength": 1,
"maxLength": 1200
},
"suggested_outcome": {
"type": "string",
"enum": [
"AGENT_ONLY",
"HUMAN_REVIEW",
"DO_NOT_DELEGATE"
]
}
},
"required": [
"title",
"scenario",
"why_it_matters"
],
"additionalProperties": false
},
"page": "/ai-delegation-boundary-webmcp/"
}review_delegation_revisionAction
Verify Guardrails as deterministic invariants across the complete finite decision domain, run known-decision regression checks, then verify the Agent Challenge gate. An Agent cannot evade a Guardrail by omitting a dangerous scenario from its Challenges. A revision with incomplete invariant verification, zero challenges, or any unresolved challenge cannot become READY_FOR_DECISION. READY_FOR_DECISION is never human approval.
View tool JSON
{
"name": "review_delegation_revision",
"kind": "act",
"impl": "imperative",
"description": "Verify Guardrails as deterministic invariants across the complete finite decision domain, run known-decision regression checks, then verify the Agent Challenge gate. An Agent cannot evade a Guardrail by omitting a dangerous scenario from its Challenges. A revision with incomplete invariant verification, zero challenges, or any unresolved challenge cannot become READY_FOR_DECISION. READY_FOR_DECISION is never human approval.",
"inputSchema": {
"type": "object",
"properties": {},
"additionalProperties": false
},
"page": "/ai-delegation-boundary-webmcp/"
}inspect_revision_historyAnswer
Read the revision timeline, including status, author, change summary, unresolved challenges, guardrail violations, and regressions. This does not modify the workspace.
View tool JSON
{
"name": "inspect_revision_history",
"kind": "answer",
"impl": "imperative",
"description": "Read the revision timeline, including status, author, change summary, unresolved challenges, guardrail violations, and regressions. This does not modify the workspace.",
"inputSchema": {
"type": "object",
"properties": {},
"additionalProperties": false
},
"page": "/ai-delegation-boundary-webmcp/"
}