kvlar-core 0.2.0

Core policy engine for Kvlar — evaluates agent actions against security policies
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91

# Default Kvlar Security Policy
#
# A sensible starting point for securing AI agents.
# Denies dangerous operations, requires approval for sensitive ones,
# and allows safe read-only operations.

name: default
description: Default security policy for AI agents
version: "1.0"

rules:
  # --- DENY: Dangerous operations ---

  - id: deny-shell
    description: Block shell/bash command execution
    match_on:
      action_types: ["tool_call"]
      resources: ["bash", "shell", "exec", "subprocess"]
    effect:
      type: deny
      reason: "Shell command execution is not permitted"

  - id: deny-file-delete
    description: Block file deletion
    match_on:
      action_types: ["tool_call"]
      resources: ["delete_file", "remove_file", "unlink"]
    effect:
      type: deny
      reason: "File deletion is not permitted"

  - id: deny-network-write
    description: Block outbound network writes
    match_on:
      action_types: ["tool_call"]
      resources: ["http_post", "http_put", "http_delete"]
    effect:
      type: deny
      reason: "Outbound network writes are not permitted"

  # --- REQUIRE APPROVAL: Sensitive operations ---

  - id: approve-email
    description: Require approval before sending emails
    match_on:
      resources: ["send_email", "smtp"]
    effect:
      type: require_approval
      reason: "Sending emails requires human approval"

  - id: approve-file-write
    description: Require approval before writing files
    match_on:
      action_types: ["tool_call"]
      resources: ["write_file", "create_file", "append_file"]
    effect:
      type: require_approval
      reason: "Writing files requires human approval"

  - id: approve-database-write
    description: Require approval for database mutations
    match_on:
      resources: ["db_insert", "db_update", "db_delete"]
    effect:
      type: require_approval
      reason: "Database mutations require human approval"

  # --- ALLOW: Safe operations ---

  - id: allow-file-read
    description: Allow reading files
    match_on:
      action_types: ["tool_call"]
      resources: ["read_file", "list_files", "search_files"]
    effect:
      type: allow

  - id: allow-network-read
    description: Allow fetching data from the network
    match_on:
      action_types: ["tool_call"]
      resources: ["http_get", "fetch"]
    effect:
      type: allow

  - id: allow-database-read
    description: Allow database queries (read-only)
    match_on:
      resources: ["db_query", "db_select"]
    effect:
      type: allow