pantheon-opencode 2.1.4 → 2.1.6
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/AGENTS.global.md +6 -6
- package/AGENTS.md +1 -1
- package/CHANGELOG.md +52 -0
- package/README.md +60 -233
- package/README.pt-BR.md +64 -220
- package/package.json +1 -1
- package/plugin.json +1 -1
- package/pyproject.toml +1 -1
- package/scripts/install/opencode.mjs +12 -0
- package/scripts/test-slice.mjs +34 -2
- package/skills-lock.json +1 -1
- package/src/agents/aphrodite.md +3 -1
- package/src/agents/apollo.md +1 -0
- package/src/agents/athena.md +1 -0
- package/src/agents/demeter.md +2 -0
- package/src/agents/gaia.md +1 -0
- package/src/agents/hephaestus.md +1 -0
- package/src/agents/hermes.md +2 -1
- package/src/agents/iris.md +1 -0
- package/src/agents/mnemosyne.md +1 -0
- package/src/agents/nyx.md +1 -0
- package/src/agents/prometheus.md +1 -0
- package/src/agents/talos.md +1 -0
- package/src/agents/themis.md +1 -0
- package/src/agents/zeus.md +8 -2
- package/src/instructions/backend-standards.instructions.md +5 -0
- package/src/instructions/test-economy.instructions.md +6 -6
- package/src/mcp/pantheon_vision.py +4 -1
- package/src/pantheon/auto-permissions.ts +4 -3
- package/src/pantheon/deepwork/deepwork-tool.ts +23 -75
- package/src/pantheon/deepwork/plan-schema.ts +4 -2
- package/src/pantheon/deepwork/plan-tracker.ts +380 -90
- package/src/pantheon/jev-decision-tool.ts +1 -7
- package/src/pantheon/local-decision-tool.ts +1 -1
- package/src/pantheon/native-vision.ts +1 -1
- package/src/pantheon/skill-search-tool.ts +241 -47
- package/src/pantheon/tool-context.ts +23 -0
- package/src/pantheon/tools.ts +46 -41
- package/src/plugin/index.js +521 -222
- package/src/plugin.ts +1 -0
- package/src/plugins/tui/dist/tui.d.ts +7 -6
- package/src/plugins/tui/dist/tui.d.ts.map +1 -1
- package/src/plugins/tui/dist/tui.js +198 -269
- package/src/plugins/tui/dist/tui.js.map +1 -1
- package/src/plugins/tui/package-lock.json +2 -2
- package/src/plugins/tui/package.json +1 -1
- package/src/plugins/tui/src/index.tsx +7 -45
- package/src/plugins/tui/src/tui-prefs.ts +3 -9
- package/src/skills/test-slicing/SKILL.md +14 -8
package/AGENTS.global.md
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
<!-- pantheon:generated:begin — Pantheon-managed region; regenerated on install/upgrade, never edit inside -->
|
|
2
2
|
<!-- Generated by scripts/build-agents-md.mjs — do not edit directly. -->
|
|
3
3
|
<!-- Source of truth: src/instructions/*.instructions.md + the static sections in this script. -->
|
|
4
|
-
<!-- pantheon:agents-md v2 artifact=global generated-at=2026-11-
|
|
4
|
+
<!-- pantheon:agents-md v2 artifact=global generated-at=2026-11-16T06:25:36-03:00 sha=de352bf -->
|
|
5
5
|
|
|
6
6
|
# Pantheon Agent System — OpenCode
|
|
7
7
|
|
|
@@ -90,11 +90,11 @@ Required: position, reasoning, trade_offs, risks, confidence. Optional `agreemen
|
|
|
90
90
|
|
|
91
91
|
Write one focused test per behavior; do not duplicate coverage.
|
|
92
92
|
|
|
93
|
-
- **
|
|
94
|
-
- **
|
|
95
|
-
- **Failure/Fix Loops**: Rerun ONLY failed checks; never
|
|
96
|
-
- **Mapping Gaps
|
|
97
|
-
-
|
|
93
|
+
- **Risk-Based Slicing**: Run narrowest slice (<10s) first; broaden only on cross-module, contract, auth, or DB risk.
|
|
94
|
+
- **Suite Escalation**: Full suites are for CI/release; state if execution is agent recommendation or mandatory gate.
|
|
95
|
+
- **Failure/Fix Loops**: Rerun ONLY failed checks; never repeat passing tests. Never delete tests to speed up checks.
|
|
96
|
+
- **Mapping Gaps & Receipts**: Missing tests use static check plus nearest slice; record command, duration, and pending gates.
|
|
97
|
+
- **CLI vs Code**: Run CLI binaries (`opencode-quota`, `git`, `npm`) via `shell`/`bash`, not eval; on failure, never glob `~/.config`.
|
|
98
98
|
- Test economy does NOT weaken stronger gates: auth, security, data-integrity, and required coverage.
|
|
99
99
|
|
|
100
100
|
<!-- Source: src/instructions/yagni.instructions.md -->
|
package/AGENTS.md
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
<!-- pantheon:generated:begin — Pantheon-managed region; regenerated on install/upgrade, never edit inside -->
|
|
2
2
|
<!-- Generated by scripts/build-agents-md.mjs — do not edit directly. -->
|
|
3
3
|
<!-- Source of truth: project sections in scripts/build-agents-md.mjs. -->
|
|
4
|
-
<!-- pantheon:agents-md v2 artifact=project generated-at=2026-11-
|
|
4
|
+
<!-- pantheon:agents-md v2 artifact=project generated-at=2026-11-16T06:25:36-03:00 sha=de352bf -->
|
|
5
5
|
|
|
6
6
|
# Pantheon — Repository Setup
|
|
7
7
|
|
package/CHANGELOG.md
CHANGED
|
@@ -13,6 +13,58 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
|
|
|
13
13
|
`node scripts/versioning.mjs apply --notes` pre-fills the generated sections from conventional commits.
|
|
14
14
|
Running apply will move this section to a versioned entry and reset the template below. -->
|
|
15
15
|
|
|
16
|
+
## [v2.1.6] - 2026-11-16
|
|
17
|
+
|
|
18
|
+
### 🆕 What's New
|
|
19
|
+
|
|
20
|
+
- **Deepwork Resilient Normalization**: Supported multi-decision blueprints, single-option policies, and string acceptance criteria.
|
|
21
|
+
- **Fail-Fast Anti-Mocking**: Prohibited fake mock servers on service ports to hide missing dependencies; fail fast with explicit errors.
|
|
22
|
+
- **Workspace Tool Directory**: Native shell tools dynamically resolve to the active project root rather than falling back to `$HOME`.
|
|
23
|
+
- **Atomic Skill Installation**: Enforced sequential `&&` chaining for `npx skills add` to prevent silent dropping of trailing targets.
|
|
24
|
+
- **Project Bootstrap Git Init**: `init --project` initializes an empty Git repo when absent, enabling worktree and branch support.
|
|
25
|
+
- **Decoupled Background Processes**: Mandated `setsid nohup <cmd> > <log> 2>&1 &` (or `disown`) for long-running service processes.
|
|
26
|
+
|
|
27
|
+
### Fixed
|
|
28
|
+
|
|
29
|
+
- **deepwork** — fixed `readPlan` schema rejection on single-option decision policies and auto-inferred acceptance criteria kinds (`test|command|file|manual`).
|
|
30
|
+
- **deepwork** — added atomic plan locking and safe temp files preventing write race conditions during concurrent decision resolutions.
|
|
31
|
+
- **tools** — flattened JSON schema for `pantheon_jev_decide` to eliminate upstream validator rejections on top-level `oneOf`/`anyOf`.
|
|
32
|
+
- **tools** — bound `shellTool` fallback execution directory dynamically to `hostLocationDirectory`.
|
|
33
|
+
- **mcp** — improved `pantheon-vision` error reporting with status code diagnostics and actionable fallback to native `vision_describe`.
|
|
34
|
+
- **tui** — removed duplicate todo block from settings preferences in favor of native sidebar.
|
|
35
|
+
- **agents** — updated Zeus and backend instructions with Fail-Fast, resource exclusivity, and decoupled server execution standards.
|
|
36
|
+
|
|
37
|
+
## [v2.1.5] - 2026-11-14
|
|
38
|
+
|
|
39
|
+
### 🆕 What's New
|
|
40
|
+
|
|
41
|
+
- **Risk-Based Validation (#444)**: Protocol distinguishes focused test slices, package suites, and mandatory repository CI gates.
|
|
42
|
+
- **Tool Execution Fallback**: Added `execute` alias executing commands via bash, preventing JS eval syntax errors on CLI commands.
|
|
43
|
+
- **CLI Execution Mandate**: Enforced CLI execution via `shell`/`bash` for binaries like `opencode-quota` with anti-panic search ban.
|
|
44
|
+
- **Scope Transparency (#444)**: `subtask_summary` explicitly reports validation scope and deferred pending repository CI pipeline gates.
|
|
45
|
+
- **Unified Skill Discovery**: Dynamic filesystem indexing of local and global skills; Zeus handoff passes explicit skill file paths.
|
|
46
|
+
|
|
47
|
+
### Added
|
|
48
|
+
|
|
49
|
+
- **tools** — added `execute` fallback alias executing workspace commands and scripts via `/bin/bash` in `src/pantheon/tools.ts`.
|
|
50
|
+
- **permissions** — whitelisted `opencode-quota` in `SAFE_DEV_BINARIES` and installer bash permission rules.
|
|
51
|
+
- **skills** — added dynamic filesystem scanner for `.opencode/skills` and `.agents/skills` returning real path and scope.
|
|
52
|
+
|
|
53
|
+
### Changed
|
|
54
|
+
|
|
55
|
+
- **tests (#444)** — enhanced `test-slice.mjs` with risk escalation triggers, validation scope reporting, and pending CI gate receipts.
|
|
56
|
+
- **instructions (#444)** — updated `test-economy` and `test-slicing` with risk-based validation protocol, anti-panic rules, and CLI mandate.
|
|
57
|
+
- **agents** — updated Zeus and specialist agent prompts in `roster.json` to mandate `SKILL.md` inspection before implementation.
|
|
58
|
+
|
|
59
|
+
### ✅ Closed Issues
|
|
60
|
+
|
|
61
|
+
- #444 - define risk-based validation protocol for pantheon agents
|
|
62
|
+
|
|
63
|
+
<!-- Add new changes here.
|
|
64
|
+
Canonical sections (docs/RELEASING.md): ### 🆕 What's New (curated highlights, ≤6 bullets ≤140 chars) →
|
|
65
|
+
### Added / Changed / Deprecated / Removed / Fixed / Security → ### ⚠️ Known Issues / ### ✅ Closed Issues.
|
|
66
|
+
`node scripts/versioning.mjs apply --notes` pre-fills the generated sections from conventional commits.
|
|
67
|
+
Running apply will move this section to a versioned entry and reset the template below. -->
|
|
16
68
|
## [v2.1.4] - 2026-11-14
|
|
17
69
|
|
|
18
70
|
### 🆕 What's New
|
package/README.md
CHANGED
|
@@ -1,268 +1,95 @@
|
|
|
1
1
|
# Pantheon
|
|
2
2
|
|
|
3
|
-
|
|
4
|
-
|
|
5
|
-
|
|
6
|
-
and teams who want useful structure without giving up speed or control of their code.
|
|
3
|
+
A multi-agent orchestration plugin for OpenCode that carries a request from plan to
|
|
4
|
+
reviewed change with 14 specialized agents, enforced TDD and explicit human approval
|
|
5
|
+
gates — for developers and teams who want structure without giving up control of the code.
|
|
7
6
|
|
|
8
|
-
[Português (Brasil)](README.pt-BR.md) ·
|
|
9
|
-
[Repository](https://github.com/ils15/pantheon-opencode) · [MIT License](LICENSE)
|
|
7
|
+
[Português (Brasil)](README.pt-BR.md) · [Repository](https://github.com/ils15/pantheon-opencode) · [MIT License](LICENSE)
|
|
10
8
|
|
|
11
|
-
[](LICENSE)
|
|
10
|
+
[](package.json)
|
|
11
|
+
[](https://opencode.ai/)
|
|
12
12
|
[](https://ils15.github.io/pantheon-opencode/)
|
|
13
|
-
[](https://github.com/ils15/pantheon-opencode/actions)
|
|
14
|
-
[](https://doi.org/10.5281/zenodo.22650136)
|
|
15
13
|
|
|
16
|
-
## What is
|
|
14
|
+
## What it is
|
|
17
15
|
|
|
18
|
-
|
|
19
|
-
|
|
20
|
-
|
|
21
|
-
|
|
16
|
+
One agent planning, writing, testing, reviewing and documenting at the same time gives you
|
|
17
|
+
fragmented context and skipped tests. Pantheon splits that work across specialists that
|
|
18
|
+
share a single workflow: Zeus orchestrates, Athena plans, the implementation agents write,
|
|
19
|
+
Themis reviews, and every phase stops for your approval. Bounded micro-edits skip the
|
|
20
|
+
ceremony and verify in seconds; heavy multi-phase work runs as a checkpointed plan that
|
|
21
|
+
resumes across sessions. It is an OpenCode plugin plus installer — no separate runtime, no
|
|
22
|
+
auto-merge, no silent permission changes.
|
|
22
23
|
|
|
23
|
-
|
|
24
|
-
|
|
25
|
-
|
|
26
|
-
|
|
27
|
-
Z -->|"ambiguous route"| J["Jev decision router<br/>typed · budgeted · reversible"]
|
|
28
|
-
J -->|"orders · ranks · verifies"| S
|
|
29
|
-
S -->|"last resort"| M["Native MCP<br/>memory · persistence · vision<br/>code-mode · resources"]
|
|
30
|
-
Z --> G["J1 risk gate<br/>off / shadow / warn / enforce"]
|
|
31
|
-
G -->|"high risk"| H["Human approval"]
|
|
32
|
-
```
|
|
33
|
-
|
|
34
|
-
**The escalation ladder every agent follows**: session context first → free local selector → one typed Jev decision when genuinely ambiguous → native MCP call only as the last resort for state the session cannot reconstruct.
|
|
35
|
-
|
|
36
|
-
```mermaid
|
|
37
|
-
flowchart LR
|
|
38
|
-
R["rough request"] --> T0{"clear &<br/>bounded?"}
|
|
39
|
-
T0 -->|yes| X["execute directly"]
|
|
40
|
-
T0 -->|"vague /<br/>multi-concern"| T1["Tier 1 — internal<br/>minimal adjust ≤15 lines<br/>(silent, no approval)"]
|
|
41
|
-
T1 --> T2q{">> given?"}
|
|
42
|
-
T2q -->|yes| T2["Tier 2 — full contract<br/>6 sections, confirm-or-adjust"]
|
|
43
|
-
T2q -->|no| X
|
|
44
|
-
```
|
|
45
|
-
|
|
46
|
-
**Skill discovery runs a trust cascade** — canonical first, popularity never compensates for provenance:
|
|
47
|
-
|
|
48
|
-
```mermaid
|
|
49
|
-
flowchart LR
|
|
50
|
-
L["installed pool<br/>(always first)"] -->|"gap"| C1["anthropics/skills<br/>canonical"]
|
|
51
|
-
C1 -->|"no fit"| C2["skills.sh<br/>install-count signal"]
|
|
52
|
-
C2 -->|"no fit"| C3["GitHub raw<br/>widest, weakest"]
|
|
53
|
-
C3 --> J["Jev orders<br/>the candidates"]
|
|
54
|
-
J --> P["YOU review<br/>and run the install"]
|
|
55
|
-
```
|
|
24
|
+
Every dispatch carries a concise handoff brief — objective, bounded scope, relevant inputs,
|
|
25
|
+
constraints, expected return format and verification — and work is routed **qualitatively**,
|
|
26
|
+
by uncertainty, domain expertise, dependencies, risk and verification, not by fixed file or
|
|
27
|
+
line thresholds.
|
|
56
28
|
|
|
57
|
-
##
|
|
29
|
+
## Installation
|
|
58
30
|
|
|
59
|
-
|
|
60
|
-
|
|
61
|
-
- **Execution proof receipts:** Cryptographic SHA-256 fingerprinting on modified files and test slice exit codes prevents false-positive task reports.
|
|
62
|
-
- **KV-cache prefix optimization:** Strict invariant-first prompt layering maximizes LLM cache hits (up to 90% cost and latency savings).
|
|
63
|
-
- **Reactive event bus & zero polling:** Process-wide in-memory event bus delivers instant updates across jobs and agents without disk polling.
|
|
64
|
-
- **Themis local reflex loops:** Specialist self-healing resolves isolated lint and test slice discrepancies before escalating to Zeus.
|
|
65
|
-
- **Assertive council (`/pantheon`):** Parallel specialist consultations resolved decisively via domain authority and JEV multi-criteria decisions without endless debate turns.
|
|
66
|
-
- **You stay in charge:** Pantheon supports your judgment; it never silently alters permissions or runs unapproved destructive actions.
|
|
67
|
-
- **Non-destructive upgrades:** Updating Pantheon (`pantheon-opencode update`) refreshes skills, MCPs, and instructions while strictly preserving all user models, presets, permissions, and custom configs.
|
|
31
|
+
Requirements: [OpenCode](https://opencode.ai/docs/) **2.x** and Node.js **≥ 26.4.0**
|
|
32
|
+
([compatibility](docs/COMPATIBILITY.md)).
|
|
68
33
|
|
|
69
|
-
## Start in 2 minutes
|
|
70
|
-
|
|
71
|
-
Requirements: [OpenCode](https://opencode.ai/docs/) **2.x** and Node.js **>=26.4.0**. (See [Compatibility](docs/COMPATIBILITY.md)).
|
|
72
|
-
|
|
73
|
-
### 1. Install
|
|
74
|
-
|
|
75
|
-
**Option A — Quick Plugin Add (Global or Project):**
|
|
76
34
|
```bash
|
|
35
|
+
# Option A — as an OpenCode plugin (global or project)
|
|
77
36
|
opencode plugin add pantheon-opencode
|
|
78
|
-
# Or pin a specific version:
|
|
79
|
-
opencode plugin add pantheon-opencode@2.0.0-rc.1
|
|
80
|
-
```
|
|
81
37
|
|
|
82
|
-
|
|
83
|
-
```bash
|
|
38
|
+
# Option B — global CLI install with full MCP tooling
|
|
84
39
|
npm install -g pantheon-opencode@beta
|
|
85
40
|
pantheon-opencode init --yes
|
|
86
|
-
```
|
|
87
41
|
|
|
88
|
-
|
|
89
|
-
```bash
|
|
42
|
+
# Verify the install
|
|
90
43
|
pantheon-opencode doctor
|
|
91
|
-
```
|
|
92
|
-
Runs a diagnostic checking Decision Router & budget ledger accessibility, MCP initialization, Python virtual environments, Node version, permissions, and config integrity. In containerized or JupyterHub environments with group-writable homes, set `PANTHEON_ALLOW_GROUP_WRITABLE=1` if needed.
|
|
93
44
|
|
|
94
|
-
|
|
95
|
-
When a new version is released:
|
|
96
|
-
```bash
|
|
45
|
+
# Later: update without losing models, presets or permissions
|
|
97
46
|
pantheon-opencode update
|
|
98
47
|
```
|
|
99
|
-
The updater verifies package identity, pulls the latest release, and runs a non-destructive configuration refresh. **Your custom models, active preset, custom permissions, plugin options, and user-defined AGENTS.md sections are preserved verbatim.**
|
|
100
|
-
|
|
101
|
-
---
|
|
102
48
|
|
|
103
|
-
##
|
|
104
|
-
|
|
105
|
-
Pantheon balances speed and rigor through three distinct execution modes:
|
|
106
|
-
|
|
107
|
-
```mermaid
|
|
108
|
-
flowchart TD
|
|
109
|
-
Req["User Request"] --> Size{"Scope / Risk"}
|
|
110
|
-
Size -->|"≤ 2 files · ≤ 50 lines<br/>routine fix / typo / CSS"| Fast["Fast-Track<br/>(Sub-10s · Inline Test Slicing)"]
|
|
111
|
-
Size -->|"Architecture / Tech choice<br/>Security / Multi-stakeholder"| Council["Council Synthesis (/pantheon)<br/>(2-3 Specialists in Parallel · Domain Authority)"]
|
|
112
|
-
Size -->|"Multi-phase feature<br/>Heavy refactor · Breaking changes"| Deep["Deepwork (/pantheon-deepwork)<br/>(Athena Plan · Phase Gates · Checkpoints)"]
|
|
113
|
-
```
|
|
49
|
+
## Usage
|
|
114
50
|
|
|
115
|
-
|
|
116
|
-
For routine bug fixes, style adjustments, or localized refactors:
|
|
117
|
-
- **No planning ceremony:** Directly executes bounded changes.
|
|
118
|
-
- **Targeted test slicing:** Runs ONLY the test slice relevant to the modified file (e.g. `node --test tests/foo.test.mjs` or `pytest -k <func>`). Full repository test suites are reserved for CI and release gates.
|
|
119
|
-
- **Single-turn delivery:** Verifies and completes in under 10 seconds.
|
|
51
|
+
Inside OpenCode, dispatch the orchestrator or a workflow:
|
|
120
52
|
|
|
121
|
-
### 2. Assertive Council Synthesis (`/pantheon`)
|
|
122
|
-
When confronting architectural crossroads, technology evaluations, or security tradeoffs:
|
|
123
|
-
```text
|
|
124
|
-
/pantheon Should we use WebSockets or Server-Sent Events for live push notifications?
|
|
125
|
-
```
|
|
126
|
-
- **Parallel Dispatch:** Dispatches 2–3 specialist agents simultaneously with independent prompts.
|
|
127
|
-
- **Assertive Arbitration:** Rather than looping in repetitive debate rounds, divergences are resolved immediately via **Domain Authority Precedence**:
|
|
128
|
-
- **Security / Auth:** `@themis` has veto power over vulnerable designs.
|
|
129
|
-
- **Database / Schema:** `@demeter` carries authoritative weight on query performance and transaction safety.
|
|
130
|
-
- **Frontend / UX / A11y:** `@aphrodite` governs user interface responsiveness and accessibility.
|
|
131
|
-
- **Infrastructure / CI:** `@prometheus` governs deployment contracts and container isolation.
|
|
132
|
-
- **API / Protocols:** `@hermes` governs contract consistency and transport schemas.
|
|
133
|
-
- **Architecture / Modularity:** `@athena` governs separation of concerns.
|
|
134
|
-
- **JEV Multi-Criteria Decision:** If trade-offs balance competing requirements equally, Zeus invokes `pantheon_jev_decide` (or local decision matrix) to rank candidates deterministically in one tool call.
|
|
135
|
-
|
|
136
|
-
### 3. Checkpointed Deepwork (`/pantheon-deepwork`)
|
|
137
|
-
For multi-phase tasks spanning 10+ turns or multiple sessions:
|
|
138
53
|
```text
|
|
54
|
+
@zeus: Add CSV export to the reports page
|
|
139
55
|
/pantheon-deepwork "Implement OAuth2 PKCE authentication flow"
|
|
140
56
|
```
|
|
141
|
-
- **The Plan is the Artifact:** The canonical ground truth is stored in `.pantheon/deepwork/<slug>/plan.json` with an append-only event log (`journal.ndjson`) for resilient crash recovery.
|
|
142
|
-
- **Native OpenCode Gates (Strictly No HTML):** Approvals (Gate 1) and decision resolutions use OpenCode's native interactive `question` tool directly in the chat UI.
|
|
143
|
-
- **Dependency DAG:** Automatically projects task dependencies into live Mermaid diagrams (`flowchart LR`) inside `PLAN.md` and chat status.
|
|
144
|
-
- **Fail-First Acceptance:** Every task defines concrete acceptance criteria (`kind: test|command|file|manual`, `fails_now: true`) verified with cryptographic receipts.
|
|
145
|
-
- **Tool Automation (`pantheon_deepwork`):** Structured tool for `init`, `status`, `update_task`, `verify_acceptance`, and `resolve_decision`.
|
|
146
|
-
- **Resumability:** Checkpoints persisted to `.pantheon/deepwork/<slug>/`. Resume anytime with `/pantheon-deepwork --resume <slug>`.
|
|
147
|
-
|
|
148
|
-
---
|
|
149
|
-
|
|
150
|
-
## The 14 Specialized Agents
|
|
151
|
-
|
|
152
|
-
For a complete breakdown of permissions, tools, MCP access, and boundary constraints per agent, see the [Agent Capability Matrix](docs/AGENT-CAPABILITY-MATRIX.md).
|
|
153
57
|
|
|
154
|
-
|
|
58
|
+
`@zeus` routes to the specialists; `/pantheon` arbitrates architectural choices across
|
|
59
|
+
specialists; `/pantheon-deepwork` runs a multi-phase plan with checkpoints and gates. You
|
|
60
|
+
approve each gate and you run every commit.
|
|
155
61
|
|
|
156
|
-
|
|
157
|
-
|---|---|---|
|
|
158
|
-
| **@zeus** | Central Orchestrator | Directs workflow, manages delegations, arbitrates council decisions, enforces token efficiency. Never implements directly. |
|
|
159
|
-
| **@athena** | Architecture & Planning | Research-first planning, system boundaries, LTS policy, dependency analysis. Plan-only; never implements. |
|
|
160
|
-
| **@hermes** | Backend & APIs | FastAPI, Python async, Node/TypeScript backends, API contracts, protocol conformance. |
|
|
161
|
-
| **@aphrodite** | Frontend & UX | React 19, strict TypeScript, responsive design, WCAG accessibility, CSS optimization. |
|
|
162
|
-
| **@demeter** | Database & Persistence | Schema migrations, query optimization, N+1 query prevention, transaction integrity. |
|
|
163
|
-
| **@themis** | Security & Quality Gate | OWASP audits, static analysis (Biome/ruff), authorization gates, dead code removal. Reviews material changes. |
|
|
164
|
-
| **@prometheus** | DevOps & Infrastructure | Docker, CI/CD pipelines, runtime configuration, provider abstraction, telemetry infra. |
|
|
165
|
-
| **@talos** | Hotfix Fast Lane | Rapid, standalone fixes for small bugs, typos, and minor logic without subagent overhead. |
|
|
166
|
-
| **@apollo** | Investigation Scout | Read-only discovery: parallel codebase and web searches. Returns structured findings. |
|
|
167
|
-
| **@hephaestus**| AI Pipelines & RAG | Embeddings, vector stores, prompt engineering, agent pipelines. |
|
|
168
|
-
| **@gaia** | Geospatial & Sensing | Remote sensing, SAR, geospatial analysis, ML/DL raster classification (read-only). |
|
|
169
|
-
| **@iris** | Git & Release Ops | Branching, PRs, issue triage, release notes. Never pushes or merges without human approval. |
|
|
170
|
-
| **@mnemosyne** | Project Memory Bank | Maintains `.pantheon/memory-bank/` and records ADRs when explicitly requested. |
|
|
171
|
-
| **@nyx** | Observability & Telemetry| OpenTelemetry tracing, token/cost monitoring, execution profiling. |
|
|
62
|
+
## Structure
|
|
172
63
|
|
|
173
|
-
|
|
174
|
-
|
|
175
|
-
|
|
176
|
-
|
|
177
|
-
|
|
178
|
-
|
|
179
|
-
-
|
|
180
|
-
|
|
181
|
-
|
|
182
|
-
|
|
183
|
-
|
|
184
|
-
- `pantheon-vision`: Image analysis and diagnostic visual inspection.
|
|
185
|
-
- **Advisory JEV Decision Router:** Evaluates candidate options (`pantheon_jev_decide`), orders trade-offs, and provides formal multi-criteria decision making without modifying permissions.
|
|
186
|
-
|
|
187
|
-
---
|
|
188
|
-
|
|
189
|
-
## Settings & TUI Window (`/pantheon-settings`)
|
|
190
|
-
|
|
191
|
-
Run `/pantheon-settings` (or `Pantheon: Settings` in the command palette) to open the interactive settings window:
|
|
192
|
-
|
|
193
|
-
- **Gates & Overrides:** Toggle named behaviors live; persists to `pantheon-overrides.json`.
|
|
194
|
-
- **Model Presets:** Switch active model tier globally or per project.
|
|
195
|
-
- **Decision Model Selector:** Interactive selector to cycle across all remote decision models (`openrouter-jev`, `typesafe-jev`, `openai-decisions`, `vercel-gateway-jev`, etc.) directly from the UI.
|
|
196
|
-
- **Bilingual Settings (PT-BR / EN):** Full visual localization for Decision, Autonomy, and Safety tabs, auto-detecting environment locale with manual toggle in the TUI tab.
|
|
197
|
-
- **Updates:** Check registry channel (`beta` / `stable`) and trigger upgrades.
|
|
198
|
-
- **Doctor:** Run inline system health diagnostics.
|
|
199
|
-
- **TUI Personalization:** Adjust spacing (`comfortable` vs `compact`), separators, gauge bar styles, context display rows, and language.
|
|
200
|
-
- **Non-Destructive Overlay & Mouse Navigation:** Opens as a floating dialog overlay over the active session without blanking the workspace (configurable via `dialog` vs `fullscreen`), with full mouse click support for tabs, sidebar, rows, value cyclers, and actions.
|
|
201
|
-
- **Modernized What's New Window:** Card layout with rounded border, pagination buttons, and a dedicated `[OK, Entendido]` button with keyboard (`Enter`/`Esc`/`Q`) and mouse click support.
|
|
202
|
-
|
|
203
|
-
---
|
|
204
|
-
|
|
205
|
-
## Documentation & Tutorials
|
|
206
|
-
|
|
207
|
-
> 📖 **Official Documentation Portal:** Explore full searchable guides, interactive Mermaid diagrams, and versioned releases at **[https://ils15.github.io/pantheon-opencode/](https://ils15.github.io/pantheon-opencode/)**.
|
|
208
|
-
|
|
209
|
-
| Guides & Tutorials | Architecture & Deep Dives | Reference & Operations |
|
|
210
|
-
|---|---|---|
|
|
211
|
-
| [First Project (End-to-End)](docs/guides/first-project.md) | [Architecture](docs/ARCHITECTURE.md) | [MCP Tools](docs/mcp-tools.md) |
|
|
212
|
-
| [Agent Capability Matrix](docs/AGENT-CAPABILITY-MATRIX.md) | [Delegation & Gates](docs/PERMISSIONS-QUALITY-DELEGATION.md) | [Native Decisions (Jev)](docs/reference/native-decisions.md) |
|
|
213
|
-
| [Deepwork Blueprint](docs/guides/deepwork.md) | [Memory System](docs/MEMORY.md) | [Model Routing](docs/reference/model-routing.md) |
|
|
214
|
-
| [Pantheon Vision](docs/guides/vision.md) | [OpenCode Integration](docs/opencode.md) | [Changelog](CHANGELOG.md) |
|
|
215
|
-
| [Cookbook](docs/guides/cookbook.md) | [Compatibility](docs/COMPATIBILITY.md) | [Wiki](wiki/Home.md) |
|
|
216
|
-
|
|
217
|
-
---
|
|
218
|
-
|
|
219
|
-
## Real-World Examples
|
|
220
|
-
|
|
221
|
-
Working runnable reference projects demonstrating Pantheon orchestration in practice:
|
|
222
|
-
|
|
223
|
-
* **[SaaS Performance Dashboard](examples/saas-dashboard/) (`examples/saas-dashboard`):** Fullstack web application with React 19, Fastify backend, streaming CSV export, automated tests, and a real Deepwork execution plan (`plan.json`, `journal.ndjson`, `PLAN.md`).
|
|
224
|
-
* **[Repo Health CLI](examples/cli-tool/) (`examples/cli-tool`):** Sub-second diagnostic CLI tool demonstrating the fast-path execution of `@talos` and `@hermes`.
|
|
225
|
-
|
|
226
|
-
---
|
|
227
|
-
|
|
228
|
-
## What's New in v2.1.4
|
|
229
|
-
|
|
230
|
-
- **Global Install Shadowing & PATH Hygiene (#424):** Doctor check (H2b) detects multiple global binaries, dangling symlinks, and prefix skew. Version skew validation prevents `_npx` from binding to older global packages.
|
|
231
|
-
- **Strict Test Slice Ordering (#442):** Reordered `--test-name-pattern` flag ahead of the test file path in `scripts/test-slice.mjs` so Node's native test runner filters tests during failure/fix loops.
|
|
232
|
-
- **Security Boundaries on Internal Folders (#441):** Hardened `auto-permissions` and `zeusReadGuard` to exclude `.git/` and `node_modules/` from autonomous reads.
|
|
233
|
-
- **Universal Session Todo Tool Availability (#439):** Granted `todowrite` and `todoread` across all 14 specialist agent manifests.
|
|
234
|
-
- **Session Auto-Discovery for Zen Vision (#440):** Automated active session discovery from environment (`OPENCODE_SESSION_ID`) and SQLite `opencode.db` (`session_v2`) for `vision_describe`.
|
|
235
|
-
|
|
236
|
-
---
|
|
237
|
-
|
|
238
|
-
## What's New in v2.0.1
|
|
239
|
-
|
|
240
|
-
- **Bilingual TUI Settings (PT-BR / EN):** Visual localization for Decision Router, Autonomy & Budgets, and J1 Safety Gate tabs, with intelligent system locale auto-detection (`LANG` / `LC_ALL`) and manual preference cycling (`auto`, `pt-BR`, `en`).
|
|
241
|
-
- **Interactive Decision Model Selector:** Permanently accessible selector row in `/pantheon-settings` to cycle across all 6 remote decision models without manual JSON configuration.
|
|
242
|
-
- **Modernized Council Synthesis (`/pantheon`):** Rich synthesis artifacts featuring Domain Authority breakdown, Mermaid arbitration diagrams, and immediate action checklists.
|
|
243
|
-
- **Modernized Strategic Planning (`/pantheon-deepwork`):** Athena plans now enforce mandatory sub-10s test slicing, execution dependency diagrams, blast radius impact matrix, and rollback blueprints.
|
|
244
|
-
- **Demystified J1 Safety Ladder:** Plain-language descriptions for J1 execution modes (`shadow`, `enforce`, `off`), making telemetry and protection gates clear.
|
|
245
|
-
|
|
246
|
-
---
|
|
247
|
-
|
|
248
|
-
## What's New in v2.0.0-rc.1
|
|
249
|
-
|
|
250
|
-
- **Release Candidate Series Inauguration:** Production-ready multi-agent orchestration for OpenCode 2.x.
|
|
251
|
-
- **Agent Token Compaction (-44.3%):** Removed duplicated backend code in Apollo/Prometheus, saving 8,150 tokens per turn (~29.3M tokens/month).
|
|
252
|
-
- **Pragmatism & Anti-Pedantry Directive:** Safe diagnostic operations are authorized without bureaucratic gates, ending agent sluggishness.
|
|
253
|
-
- **Batched Suite TDD & Test-Slicing:** Cohesive multi-endpoint testing and sub-100ms targeted test execution.
|
|
254
|
-
- **Proactive Skill Discovery & Query Caching:** Cached skill lookups (24h TTL) and automated recommendation for external tools.
|
|
255
|
-
- **Live JEV Scorecard & Sandbox Web App:** 16/16 pass rate ($0.0000238/decision) verified with an end-to-end sandbox application.
|
|
256
|
-
|
|
257
|
-
See [CHANGELOG.md](CHANGELOG.md) for full historical release notes.
|
|
64
|
+
```
|
|
65
|
+
src/agents/ 14 canonical agent definitions
|
|
66
|
+
src/skills/ skills loaded on demand
|
|
67
|
+
src/mcp/ local MCP servers (memory, persistence, resources, code-mode, vision)
|
|
68
|
+
src/plugins/ OpenCode plugin + TUI
|
|
69
|
+
scripts/ installer, doctor, test-slice, release tooling
|
|
70
|
+
bin/ pantheon-opencode CLI
|
|
71
|
+
docs/ documentation (also published as the docs portal)
|
|
72
|
+
examples/ runnable reference projects
|
|
73
|
+
tests/ Node and TypeScript test suites
|
|
74
|
+
```
|
|
258
75
|
|
|
259
|
-
##
|
|
76
|
+
## Documentation
|
|
260
77
|
|
|
261
|
-
|
|
78
|
+
| Start here | Go deeper |
|
|
79
|
+
|---|---|
|
|
80
|
+
| [Docs portal](https://ils15.github.io/pantheon-opencode/) | [Architecture](docs/ARCHITECTURE.md) |
|
|
81
|
+
| [Quick Start](docs/QUICKSTART.md) | [Overview and routing ladders](docs/OVERVIEW.md) |
|
|
82
|
+
| [Installation](docs/INSTALLATION.md) | [Execution modes](docs/guides/execution-modes.md) |
|
|
83
|
+
| [First project end-to-end](docs/guides/first-project.md) | [Agent Capability Matrix](docs/AGENT-CAPABILITY-MATRIX.md) |
|
|
84
|
+
| [Settings & TUI window](docs/SETTINGS.md) | [Permissions, gates & delegation](docs/PERMISSIONS-QUALITY-DELEGATION.md) |
|
|
85
|
+
| [v2.x highlights](docs/releases/v2-highlights.md) | [CHANGELOG](CHANGELOG.md) |
|
|
262
86
|
|
|
263
|
-
|
|
87
|
+
Runnable examples: [SaaS performance dashboard](examples/saas-dashboard/) and
|
|
88
|
+
[repo health CLI](examples/cli-tool/). Contributing:
|
|
89
|
+
[CONTRIBUTING.md](CONTRIBUTING.md).
|
|
264
90
|
|
|
265
|
-
|
|
266
|
-
Cite via Zenodo concept DOI: [10.5281/zenodo.22650136](https://doi.org/10.5281/zenodo.22650136). Citation metadata is available in [CITATION.cff](CITATION.cff).
|
|
91
|
+
## License
|
|
267
92
|
|
|
268
|
-
|
|
93
|
+
MIT — see [LICENSE](LICENSE). Cite via Zenodo concept DOI
|
|
94
|
+
[10.5281/zenodo.22650136](https://doi.org/10.5281/zenodo.22650136)
|
|
95
|
+
([CITATION.cff](CITATION.cff)).
|