LocalSetup/ls/docs/bootstrap-packs/opencode-agent-team/metadata.yaml at main · CruxExperts/LocalSetup · GitHub

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
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
schema_version: 1
id: opencode-agent-team
display_name: OpenCode Agent Team Bootstrap
status: audit-ready
version: "1.1"
primary_platform: opencode
related_platforms:
  - codex
  - kilo
  - cursor
  - claude-code
  - openclaw
localsetup_pack: bootstrap
description: Generic OpenCode controller and subagent bootstrap pack with OpenCode-native agent, permission, model-slot, audit, and validation guidance.
verified_on: "2026-05-29"
source_of_truth:
  skill_workflow_pack_membership: ls/config/pack.yaml
  generated_skill_workflow_pack_index: ls/docs/_generated/skill-packs.md
  document_bundle_index: ls/docs/bootstrap-packs/INDEX.md
  docs_index: ls/docs/bootstrap-packs/INDEX.md
  readme: ls/docs/bootstrap-packs/opencode-agent-team/README.md
  audit_prompt: ls/docs/bootstrap-packs/opencode-agent-team/AUDIT_PROMPT.md
  model_map: ls/docs/bootstrap-packs/opencode-agent-team/MODEL_MAP.md
native_opencode_surfaces:
  global:
    - ~/.config/opencode/opencode.jsonc
    - ~/.config/opencode/AGENTS.md
    - ~/.config/opencode/agents/*.md
  repo:
    - AGENTS.md
    - opencode.jsonc
    - .opencode/agents/*.md
    - ls/templates/opencode/AGENTS.md
model_slots:
  Agent-Frontier: Highest-capability reasoning slot available to the installation for architecture, security, final review, and high-risk decisions.
  Agent-Main: General-purpose controller and build slot suited to integrated planning and implementation.
  Agent-Coder: Coding-capable slot for bounded implementation with exact write scope and focused checks.
  Agent-Scout: Fast, lower-cost slot for read-only exploration, research, and validation summaries.
  Agent-Lowcost: Least-cost slot that meets the quality needed for low-risk summaries, titles, and routine transformations.
opencode_role_map:
  controller:
    mode: primary
    model_slot: Agent-Main
    permissions:
      task: Allow only known subagents with wildcard deny first and explicit allows after it.
  plan:
    mode: primary
    model_slot: Agent-Main
    permissions:
      edit: deny
      bash: deny_or_restrict
  explore:
    mode: subagent
    model_slot: Agent-Scout
    permissions:
      edit: deny
      task: deny
  researcher:
    mode: subagent
    model_slot: Agent-Scout
    permissions:
      edit: deny
      task: deny
      webfetch: allow_when_needed
      websearch: allow_when_needed
  worker:
    mode: subagent
    model_slot: Agent-Coder
    permissions:
      task: deny_unless_explicitly_needed
  tester:
    mode: subagent
    model_slot: Agent-Scout
    permissions:
      edit: deny
      task: deny
  reviewer:
    mode: subagent
    model_slot: Agent-Frontier
    permissions:
      edit: deny
      bash: deny
      task: deny
current_fact_sources:
  official_docs:
    - https://opencode.ai/docs/agents/
    - https://opencode.ai/docs/config/
    - https://opencode.ai/docs/models/
    - https://opencode.ai/docs/providers/
    - https://opencode.ai/docs/permissions/
  schema:
    - https://opencode.ai/config.json
  source_snapshot:
    repository: https://github.com/anomalyco/opencode
    branch: dev
    commit: b2a06351b545dbefa30181016696ca25110b2366
  local_cli:
    opencode_version: 1.15.11
    commands:
      - opencode --version
      - opencode agent --help
      - opencode agent list
schema_facts:
  agent_field: agent
  mode_values:
    - primary
    - subagent
    - all
  default_agent: Must be a primary agent; falls back to build if unset or invalid.
  permission_task: Controls Task-tool subagent invocation by matching subagent type.
  deprecated_tools_field: AgentConfig tools is deprecated; new examples should use permission.
  provider_shape:
    npm: "@ai-sdk/openai-compatible"
    options:
      - baseURL
      - apiKey
    models: map of model ids to model config.
approval_boundaries:
  - home-directory writes
  - ~/.config/opencode changes
  - provider key or credential changes
  - global OpenCode permission changes
  - sibling-repo writes
  - destructive replacement of legacy skills or prompts
operating_defaults:
  controller: Own requirements, task decomposition, state tracking, verification, and final acceptance while defaulting non-trivial work to subagent-first routing.
  direct_mode: Use only for trivial one-step tasks, simple questions, cases with no useful independent subtask, or active tool/mode constraints.
  delegation: Use OpenCode subagents through Task-tool routing when a bounded subtask preserves context, reduces risk, or improves verification.
  normal_fanout: 1-2 agents for non-trivial tasks; use 3 only for clearly independent discovery, research, or validation.
  no_write_modes: In planning-only contexts, plan ledgers and subtasks but do not create or edit state until writes are allowed.
validation_commands:
  - opencode --version
  - opencode agent --help
  - opencode agent list
  - curl -fsSL https://opencode.ai/config.json
  - uv run --locked python ls/tools/localsetup.py --source-root . validate-catalog
  - uv run --locked python ls/tools/localsetup.py --source-root . plan --packs bootstrap
  - uv run --locked python ls/tools/generate_docs_artifacts.py --repo-root .
  - uv run --locked python ls/tools/localsetup.py --source-root . generate-docs
  - git diff --check