agy-squad 0.1.0 → 0.1.1

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.
Files changed (18) hide show
  1. package/.agents/skills/swarm-governance/references/intake-and-planning.md +1 -1
  2. package/CHANGELOG.md +10 -0
  3. package/README.md +83 -118
  4. package/bin/agy-squad.mjs +2 -2
  5. package/docs/benchmarks.md +59 -7
  6. package/docs/core-workflows.md +1 -1
  7. package/docs/getting-started.md +2 -2
  8. package/docs/index.md +7 -7
  9. package/docs/reviews/v0.1.1-review.md +397 -0
  10. package/docs/tools.md +3 -2
  11. package/mcp/briefs.mjs +13 -9
  12. package/mcp/dispatch.mjs +79 -16
  13. package/mcp/index.js +40 -3
  14. package/mcp/planner.mjs +13 -7
  15. package/package.json +2 -2
  16. package/plugin.json +1 -1
  17. package/scripts/install-register.mjs +2 -2
  18. package/scripts/release-preflight.mjs +35 -6
package/.agents/skills/swarm-governance/references/intake-and-planning.md CHANGED
@@ -48,7 +48,7 @@ When launching `/swarm` and any of the four execution choices below are unspecif
48
48
 
49
49
  ## Planner and Dispatch Commands
50
50
 
51
- `Agy-Squad` is a workspace-private package (`"private": true`), so always invoke the CLI from the workspace root with `node bin/agy-squad.mjs` (legacy `agy-squad run` and the standalone `node mcp/handoffs.mjs` CLI are retired; `mcp/handoffs.mjs` is an internal JS library only).
51
+ `Agy-Squad` can be invoked either over MCP (`squad_planner` / `dispatch_squad` — pass `workspace: "<repo-root>"` when running from a global plugin install so relative `manifestPath` resolves to the active repository) or from the workspace root via `node bin/agy-squad.mjs` (legacy `agy-squad run` and the standalone `node mcp/handoffs.mjs` CLI are retired; `mcp/handoffs.mjs` is an internal JS library only).
52
52
 
53
53
  ### Manifest & Task Planning (`squad_planner` / `agy-squad plan`)
54
54
 
package/CHANGELOG.md CHANGED
@@ -5,6 +5,16 @@ All notable changes to `Agy-Squad` will be documented in this file.
5
5
  The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
6
6
  and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
7
7
 
8
+ ## [0.1.1] - 2026-11-15
9
+
10
+ ### Fixed
11
+ - **MCP Workspace Resolution (`mcp/index.js`, `mcp/dispatch.mjs`, `mcp/planner.mjs`, `docs/tools.md`)**: Added optional `workspace` parameter on both `dispatch_squad` and `squad_planner`, forwarded explicit `workspace` overrides in `all_l1` lead delegation instructions, captured client `initialize` workspace roots (`workspaceFolders`/`rootUri`/`rootPath` with resilient URI fallback), supported absolute `manifestPath` inside symlinked workspaces, trimmed `manifestPath`/`AGY_SQUAD_WORKSPACE` consistently, and surfaced actionable `ENOENT` diagnostics (`workspace:`, `manifestPath:`, and `Pass 'workspace'` / `AGY_SQUAD_WORKSPACE` hints) even on missing default manifests when the MCP server starts outside the active repository root.
12
+ - **Bounded File-Read Guidance (`mcp/briefs.mjs`, `mcp/planner.mjs`)**: Added concise `rtk read` / `rtk grep` / bounded `<= 150`-line `read_file` (`StartLine`/`EndLine`) instructions across compiled working specialist, squad lead, and auto-synthesized `integrity_auditor` (`audit-<id>`) prompts while removing duplicated auditor text and staying inside prompt byte budgets (`1702 B <= 1800 B` specialist, `2881 B <= 3300 B` lead, `3075 B <= 4096 B` `tools/list`).
13
+ - **Release Preflight & Package Checks (`scripts/release-preflight.mjs`, `tests/verify-package.mjs`, `tests/test-server.mjs`)**: Added `mcp/dispatch.mjs`, `self-diagnosis` runtime files, `verify:package` script checks, `./` + `!` negation handling in `files`, `LICENSE` + nested scratch/log tarball guards, isolated `AGY_SQUAD_WORKSPACE` in MCP server tests, and bumped catch version fallbacks to `0.1.1`.
14
+
15
+ ### Documentation
16
+ - **Matched-`High` Benchmark Retest (`benchmarks/teamwork-vs-squad/`, `docs/benchmarks.md`, `docs/index.md`, `docs/getting-started.md`, `docs/core-workflows.md`, `README.md`, `package.json`, `zensical.toml`)**: Recorded the November 15, 2026 `Gemini 4 Hydrogen (High)` matched-effort comparison (`6m 30s` / `390s` vs. `22m 41s` / `1,361s`, `178` vs. `655` combined steps, `336.8 KB` vs. `801.8 KB` combined raw transcripts, `1,629 B` vs. `18,322 B` parent wakeup bytes, `21` vs. `16` defects, and `audit-ab-audit` `CLEAN` verdict vs. `/teamwork` Sentinel-fabricated snippets passing `Victory Auditor`) alongside the historical November 8, 2026 Phase 1 baseline, extended `verify-benchmark.mjs` to validate per-row KB sums, `N1`–`N6` rubric mappings, digest/SHA-256 provenance, and flat `/swarm` planning, scoped platform support to Linux and WSL2, and standardized GitHub Pages URL casing (`https://paragon-ux.github.io/Agy-Squad/`).
17
+
8
18
  ## [0.1.0] - 2026-11-15
9
19
 
10
20
  ### Added
package/README.md CHANGED
@@ -1,157 +1,122 @@
1
1
  # Agy-Squad
2
2
 
3
- Agy-Squad adds scoped cohort planning, role-separated auditing, and SHA-256 evidence bundling to **Antigravity Desktop (IDE)** and the **Antigravity CLI (`agy`)** on **Linux and WSL2 (macOS support is best-effort and untested)**.
3
+ [![npm version](https://img.shields.io/npm/v/agy-squad.svg)](https://www.npmjs.com/package/agy-squad)
4
+ [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
5
+ [![Docs](https://img.shields.io/badge/docs-Zensical-purple.svg)](https://paragon-ux.github.io/Agy-Squad/)
6
+ [![Tests](https://img.shields.io/badge/tests-100%25%20passing-brightgreen.svg)](https://github.com/paragon-ux/Agy-Squad/actions)
4
7
 
5
- `squad_planner` (and `dispatch_squad` for `invoke_subagent`) assigns each worker a workspace scope and a file-backed report slot (`scratch/agy-squad/.../report.md`). Specialists write full evidence to disk and post a single-line `Pointer-Over-Wire` receipt under 500 UTF-8 bytes. A parent or squad lead checks the cited code and tests directly, pairs with a flat integrity auditor (`audit-<id>` on `/swarm` or `integrity_auditor` via slot; emits `Integrity Verdict: CLEAN`), and runs `node bin/agy-squad.mjs bundle` to record SHA-256 digests for verified inputs, specialist reports, and the audit report.
8
+ `Agy-Squad` is a multi-agent `/swarm` planner, code-review harness, and SHA-256 proof bundler (`agy-squad`) for **Google Antigravity** (`agy`) on Linux and WSL2 (`Node.js >= 22`, zero external npm dependencies). It pairs with [Agy-Context-Saver](https://github.com/paragon-ux/Agy-Context-Saver) (`>= 2.0.1`) to assign non-overlapping file scopes, keep board handoffs under `500` bytes, audit worker tool traces, and verify every report with SHA-256 digests.
6
9
 
7
10
  ![Agy-Squad, Antigravity, and Agy-Context-Saver architecture](docs/assets/architecture-overview.svg)
8
11
 
9
- The repository ships four core components:
12
+ The repository ships four components:
10
13
 
11
- - **[`swarm-governance`](.agents/skills/swarm-governance/SKILL.md) (Execution Modifier) and `bin/agy-squad.mjs`:** Role separation, `Pointer-Over-Wire` receipts, and offline CLI (`dispatch`, `plan`, `verdict`, `fingerprint`, `inspect`, `digest`, `bundle`, `verify`, `prune`) layered over Antigravity native `/swarm`.
12
- - **[`self-diagnosis`](.agents/skills/self-diagnosis/SKILL.md) (Task Spec):** Zero-subagent diagnostic profiler (`diagnose.mjs`) and three specialist audit cohorts for operator-supervised skill, runtime, and token checks.
13
- - **[`pr-review`](.agents/skills/pr-review/SKILL.md) & [`full-review`](.agents/skills/full-review/SKILL.md) (Task Specs — [Standard v1.4.0](.agents/skills/pr-review/references/pr-review-standard.md)):** Ten-category branch/PR diff review (`pr-review`) and non-diff evidence/direction triage (`full-review`), producing commit-pinned `1 ID = 1 Row` ledgers under `docs/reviews/`.
14
- - **[Evidence retention and teardown](docs/core-workflows.md#4-three-scope-evidence-retention-and-teardown-contract):** SHA-256 bundle verification, report pruning, and ledger-gated `diagnose.mjs --teardown`. Native `/swarm` owns worker lifecycle and `kill` cleanup; Phase 1 temporary CLI session registrations retired with the launcher (`L17`).
15
-
16
- ## Skill composition & tool surface map
17
-
18
- | Surface | Type & Purpose | When to Use |
19
- | --- | --- | --- |
20
- | `pr-review` / `full-review` / `self-diagnosis` | **Task Specs** (what to check & report) | Use `pr-review` for PR/branch diffs, `full-review` for non-diff triage/direction decisions, `self-diagnosis` for repo/harness health audits. |
21
- | `swarm-governance` | **Execution Modifier** (how teams execute) | Pair with any Task Spec when running on `/swarm` or teams (adds slot receipts, flat `audit-<id>`, `bin/agy-squad.mjs` bundles, and 4-check closure). Ask user once if `Flat` vs `Layered` topology is unspecified. |
22
- | `squad_planner` / `node bin/agy-squad.mjs plan` | **`/swarm` & unified planner** | Default planner for `/swarm` (`swarmAdd` + `provenanceScaffold` + `bundleRecordTemplates`); omit `squadId` to plan all manifest squads (`readOnly: true` exempts read-only slots from overlap warnings). |
23
- | `dispatch_squad` / `node bin/agy-squad.mjs dispatch` | **`invoke_subagent` compiler** | Explicit `invoke_subagent` stage compiler (`all_l1` leads, `<squadId>` specialists, or `integrity_auditor[:<target>]` audit slot; legacy `adversarial_*` modes removed). |
24
- | `mcp/handoffs.mjs` | **Internal verification library** | Imported by `bin/agy-squad.mjs` for `fingerprint`, `inspect`, `bundle`, `verify`, `prune`; standalone CLI removed in favor of `node bin/agy-squad.mjs`. |
25
-
26
- ## Documentation
27
-
28
- - [Getting started (Desktop IDE & CLI on POSIX / WSL2)](docs/getting-started.md)
29
- - [Core workflows: orchestrate, self-diagnose, review, and tear down](docs/core-workflows.md)
30
- - [Roles, skill workflows, and runtime closure checklist](docs/roles-and-workflows.md)
31
- - [Concepts and terminology](docs/concepts.md)
32
- - [Manifest and tool contract](docs/tools.md)
33
- - [Stopping agents and child processes](docs/stopping-agents.md)
34
- - [Benchmarks vs. Antigravity Teamwork](docs/benchmarks.md) ([benchmark folder](benchmarks/teamwork-vs-squad/README.md))
35
- - [WSL CLI verification and live trial](docs/plans/wsl-cli-verification.md)
36
- - [Architecture decision ledger](docs/plans/LEDGER.md)
37
- - [Native enforcement ledger](docs/plans/native-enforcement-ledger.md)
14
+ - **Cohort planner & MCP server ([`mcp/planner.mjs`](mcp/planner.mjs), [`mcp/dispatch.mjs`](mcp/dispatch.mjs), [`mcp/index.js`](mcp/index.js)):** Compiles flat (`1–3` workers) and layered (`4+` workers with Squad Leads) `/swarm` cohorts via `squad_planner` (or `dispatch_squad` for `invoke_subagent`), enforcing disjoint writable file scopes, 24-character roster names, explicit model pinning, and first-line `[agy-transcript-access: allow|deny]` directives.
15
+ - **Verification & proof CLI ([`bin/agy-squad.mjs`](bin/agy-squad.mjs), [`mcp/handoffs.mjs`](mcp/handoffs.mjs), [`mcp/verdicts.mjs`](mcp/verdicts.mjs)):** Fingerprints verified source files, reconciles `<= 500` byte `Pointer-Over-Wire` board receipts, computes canonical `Evidence Digest` hashes for independent integrity auditors (`audit-<id>`), and writes tamper-evident `.sources.json` bundles (`plan`, `dispatch`, `fingerprint`, `inspect`, `digest`, `verdict`, `bundle`, `verify`, `prune`).
16
+ - **Review & governance skills ([`.agents/skills/`](.agents/skills/)):** Installs [`pr-review`](.agents/skills/pr-review/SKILL.md) (branch/PR diffs), [`full-review`](.agents/skills/full-review/SKILL.md) (non-diff triage and direction decisions), and [`swarm-governance`](.agents/skills/swarm-governance/SKILL.md) (role separation, proof bundling, and 4-check runtime closure) using the 10-category [PR Review Standard v1.4.0](.agents/skills/pr-review/references/pr-review-standard.md).
17
+ - **Zero-subagent harness profiler ([`self-diagnosis`](.agents/skills/self-diagnosis/SKILL.md), [`diagnose.mjs`](.agents/skills/self-diagnosis/scripts/diagnose.mjs)):** Audits skill token budgets, prompt template byte ceilings (`<= 1800 B` specialist, `<= 3300 B` lead), review ledger invariants, and package boundaries in `< 200 ms` with zero model cost.
38
18
 
39
19
  ## Quickstart
40
20
 
41
- Live runs follow a six-step cycle across Antigravity `/swarm` execution and `bin/agy-squad.mjs` verification:
42
-
43
21
  ![Six-step swarm lifecycle and dual-verdict gate](docs/assets/swarm-lifecycle.svg)
44
22
 
45
23
  ### Prerequisites
46
24
 
47
- - **Operating system:** Linux or WSL2 on Windows (`Node.js >= 22`); macOS is best-effort and untested.
48
- - **Antigravity environment:** Antigravity Desktop (IDE) with `/swarm` and/or `invoke_subagent`.
49
- - **Companion governor:** `Agy-Context-Saver` (`>= 2.0.1`) reporting active status via `agy-context status` with `RTK` (`>= 1.1.0`). Run `agy-context setup` once on a new host if `rtk` or the plugin hook is missing; see [Getting started](docs/getting-started.md#host-and-companion-compatibility-matrix).
50
-
51
- ### 1. Install and register via NPX (Recommended — zero clone required)
25
+ - **Node.js:** `>= 22.0.0` (Linux or WSL2 on Windows)
26
+ - **Google Antigravity:** Antigravity 2.0+ (Desktop IDE or `agy` CLI with `/swarm` / `invoke_subagent`)
27
+ - **Agy-Context-Saver:** `>= 2.0.1` with **RTK (`>= 1.1.0`)** (`npx --package=agy-context-saver agy-context setup`)
52
28
 
53
- Run the single-command installer to register the native Antigravity plugin (`~/.gemini/config/plugins/agy-squad`), link the governance skills (`pr-review`, `full-review`, `self-diagnosis`, `swarm-governance`), register the `agy-squad` MCP server (`squad_planner`, `dispatch_squad`), and install the `agy-squad` CLI shim (`~/.gemini/antigravity/bin/agy-squad`):
29
+ ### Install and verify
54
30
 
55
31
  ```bash
56
- # Install and activate Agy-Squad (pairs with Agy-Context-Saver >= 2.0.1)
32
+ # Install via npx (registers plugin, skills, MCP server, and agy-squad CLI shim)
57
33
  npx --package=agy-squad agy-squad setup
58
34
 
59
- # Check installation health
35
+ # Or from a local checkout
36
+ npm run setup
37
+
38
+ # Check plugin, skills, MCP registration, and CLI shim status
60
39
  agy-squad status
61
40
  ```
62
41
 
63
- ### 2. Optional: Local repository checkout or manual MCP registration
64
-
65
- If you are developing `Agy-Squad` locally, clone the repository and run `npm run setup && npm test`:
42
+ To remove the plugin link, MCP server entry, and CLI shims:
66
43
 
67
44
  ```bash
68
- git clone https://github.com/paragon-ux/Agy-Squad.git
69
- cd Agy-Squad
70
- npm run setup
71
- npm test
45
+ agy-squad uninstall
46
+ # Or restore pre-install .bak configuration files:
47
+ agy-squad uninstall --restore-backups
72
48
  ```
73
49
 
74
- To configure a workspace-scoped `.agents/mcp_config.json` manually instead of using `agy-squad setup`:
75
-
76
- ```json
77
- {
78
- "mcpServers": {
79
- "agy-squad": {
80
- "command": "/absolute/path/to/node",
81
- "args": ["/absolute/path/to/Agy-Squad/mcp/index.js"],
82
- "env": {
83
- "AGY_SQUAD_WORKSPACE": "/absolute/path/to/workspace"
84
- },
85
- "tools": {
86
- "dispatch_squad": { "eager": true },
87
- "squad_planner": { "eager": true }
88
- }
89
- }
90
- }
91
- }
92
- ```
93
-
94
- Once connected, call `squad_planner` (for `/swarm` cohorts with explicit `model` pinning, `readOnly`-aware scope-overlap checks, and `L6`-safe `provenanceScaffold`) or `dispatch_squad` (`invoke_subagent`), or run `/swarm-governance`, `/pr-review`, `/full-review`, or `/self-diagnosis` in chat.
95
-
96
- ### 3. Plan and verify via the Governance CLI (`bin/agy-squad.mjs`)
97
-
98
- Run `node bin/agy-squad.mjs` from the workspace root. Legacy `agy-squad run` (`bin/terminal.py`) and the standalone `node mcp/handoffs.mjs` CLI were retired in PR #5 (`L17`, `L42`, `L43`).
50
+ ### Plan, audit, and bundle a cohort
99
51
 
100
52
  ```bash
101
- # Plan a swarm cohort with explicit model pinning and L6 provenance scaffold:
102
- node bin/agy-squad.mjs plan --manifest .agents/skills/self-diagnosis/resources/self-diagnosis-manifest.json \
103
- --model gemini-4-hydrogen-high
104
-
105
- # Or compile an invoke_subagent stage via dispatch:
106
- node bin/agy-squad.mjs dispatch --squad skill_and_pr_review_audit \
107
- --manifest .agents/skills/self-diagnosis/resources/self-diagnosis-manifest.json
108
-
109
- # Verify strict verdicts, fingerprint verified files, and bundle/verify/prune SHA-256 evidence:
110
- node bin/agy-squad.mjs verdict scratch/agy-squad/squad-<id>/lead/report.md
111
- node bin/agy-squad.mjs verdict --integrity scratch/agy-squad/squad-integrity_auditor/specialist-1/report.md # or squad-integrity_auditor-<id>/... for multi-squad
112
- node bin/agy-squad.mjs digest <bundle-record.json>
113
- node bin/agy-squad.mjs bundle <bundle-record.json>
114
- node bin/agy-squad.mjs verify scratch/agy-squad/squad-<id>/lead/report.sources.json
53
+ # Plan a layered /swarm cohort from a manifest or freeform task
54
+ agy-squad plan --manifest .agents/skills/self-diagnosis/resources/self-diagnosis-manifest.json \
55
+ --topology layered --mode swarm
56
+
57
+ # Parse strict content and integrity verdicts
58
+ agy-squad verdict scratch/agy-squad/squad-<id>/lead/report.md
59
+ agy-squad verdict --integrity scratch/agy-squad/squad-integrity_auditor-<id>/specialist-1/report.md
60
+
61
+ # Compute canonical Evidence Digest, write .sources.json proof bundle, and verify SHA-256 hashes
62
+ agy-squad fingerprint src/app.js docs/tools.md
63
+ agy-squad inspect scratch/agy-squad/<id>-record.json
64
+ agy-squad digest scratch/agy-squad/<id>-record.json
65
+ agy-squad bundle scratch/agy-squad/<id>-record.json --replace
66
+ agy-squad verify scratch/agy-squad/squad-<id>/lead/report.sources.json
115
67
  ```
116
68
 
117
- Use `manage_subagents` or `/swarm` status (`swarm{action: "status"}`) to inspect active workers or stop children ([Stopping agents](docs/stopping-agents.md)).
118
-
119
- ## Core workflows: orchestrate, diagnose, and review
120
-
121
- - **Govern native `/swarm` or dispatch tasks:** Pair Antigravity native `/swarm` and `squad_planner` with [`swarm-governance`](.agents/skills/swarm-governance/SKILL.md) (or dispatch `invoke_subagent` stages via `dispatch_squad` and flat `integrity_auditor:<target>`), collect `<= 500` B `READY` pointers, enforce strict verdicts, and bundle SHA-256 digests with `node bin/agy-squad.mjs` (`dispatch`, `plan`, `verdict`, `fingerprint`, `inspect`, `digest`, `bundle`, `verify`, `prune`) ([Core workflows](docs/core-workflows.md#1-orchestrate-proportional-specialist-or-lead-cohorts)).
122
- - **Self-diagnose the harness:** Run `node .agents/skills/self-diagnosis/scripts/diagnose.mjs --markdown` for a zero-subagent health and token-budget check, or compile targeted audit cohorts with `--print-plan` (`--print-dispatch` for subagents) ([`self-diagnosis` skill](.agents/skills/self-diagnosis/SKILL.md)).
123
- - **Review pull requests or evidence bundles:** Apply the ten-category `v1.4.0` repository review standard ([`pr-review`](.agents/skills/pr-review/SKILL.md) for diffs, [`full-review`](.agents/skills/full-review/SKILL.md) for non-diff triage/direction) and record `1 ID = 1 Row` ledgers in `docs/reviews/` before running `diagnose.mjs --teardown`.
124
-
125
- ## Positioning and benchmark summary
126
-
127
- | Approach | Work allocation | Best fit | Handoff, verification, and teardown |
128
- | --- | --- | --- | --- |
129
- | **Agy-Squad + Agy-Context-Saver (`v2.0.1+`)** | Task-proportional stage compile (`squad_planner` / `dispatch_squad`) for direct specialists or domain leads in Desktop IDE or CLI | Scoped coding or review tasks on POSIX / WSL2 needing compact receipts, SHA-256 bundles, `self-diagnosis`, `pr-review` (`v1.4.0`), and governed teardown | `<= 500` B `Pointer-Over-Wire` receipt, parent verification + SHA-256 digest check (`20 ms` in Nov 8 Phase 1 trial; current CLI `node bin/agy-squad.mjs verify`), and 3-scope teardown contract |
130
- | **Antigravity `/teamwork`** (or `/teamwork-preview`) | Autonomous wrapper, internal orchestrator, exploratory workers, and semantic review fleet | Long multi-milestone campaigns suited to autonomous planning and LLM reviewer/auditor subagents | Top-level milestone delegation with internal Sentinel updates and dedicated LLM review agents |
131
- | **Antigravity native subagents** (`invoke_subagent`) | Direct manual delegation | One-off parallel tasks without a manifest or report-bundle pipeline | Native subagent return messages |
132
-
133
- ### Recorded head-to-head audit vs. Antigravity `/teamwork`
134
-
135
- On November 8, 2026, a recorded sequential audit ran both arms against the six-defect event-relay fixture ([`benchmarks/teamwork-vs-squad/`](benchmarks/teamwork-vs-squad/README.md)). Both caught all six planted defect categories (`6 / 6`).
69
+ ## CLI, MCP, and skills summary
70
+
71
+ | Surface | Purpose | Common usage |
72
+ | :--- | :--- | :--- |
73
+ | `agy-squad setup` / `status` / `uninstall` | Install, audit, or remove plugin link, skills, MCP entry, and CLI shim | `agy-squad setup` / `agy-squad status` / `agy-squad setup --check` |
74
+ | `agy-squad plan` / `squad_planner` (MCP) | Compile flat or layered `/swarm` cohorts with scope checks and `provenanceScaffold` | `agy-squad plan --manifest <path> [--topology flat\|layered]` |
75
+ | `agy-squad dispatch` / `dispatch_squad` (MCP) | Compile `invoke_subagent` payloads (`all_l1`, `<squadId>`, `integrity_auditor[:<id>]`) | `agy-squad dispatch --squad <id> --manifest <path>` |
76
+ | `agy-squad fingerprint` / `inspect` / `digest` | Hash verified files, check `< 500 B` `READY` receipts, and compute `Evidence Digest` | `agy-squad digest scratch/agy-squad/<id>-record.json` |
77
+ | `agy-squad verdict` / `bundle` / `verify` / `prune` | Parse `PASS` / `CLEAN` verdicts, write `.sources.json` proofs, and prune loose reports | `agy-squad bundle <record.json> --replace` / `agy-squad verify <sources.json>` |
78
+ | `/pr-review` & `/full-review` (`v1.4.0`) | 10-category commit-pinned diff review (`pr-review`) or non-diff triage (`full-review`) | Pair with `/swarm-governance` on `/swarm` |
79
+ | `/self-diagnosis` (`diagnose.mjs`) | Zero-subagent profiler checking skill token budgets, prompt sizes, and review ledgers | `node .agents/skills/self-diagnosis/scripts/diagnose.mjs --markdown` |
80
+
81
+ ## Benchmark summary vs. Antigravity `/teamwork`
82
+
83
+ In a matched-effort sequential audit (`Gemini 4 Hydrogen (High)` on all parents and workers, November 15, 2026) against the six-defect event-relay fixture ([`benchmarks/teamwork-vs-squad/`](benchmarks/teamwork-vs-squad/README.md)) using native `/swarm` + `Agy-Context-Saver 2.0.1` (`rtk 1.1.0`), both arms caught all six planted defect categories (`6 / 6`):
84
+
85
+ | Metric (Parent + All Workers End-to-End) | Agy-Squad + Agy-Context-Saver (`flat /swarm`) | Antigravity `/teamwork` (`[skip-interview]`) | Difference |
86
+ | :--- | :--- | :--- | :--- |
87
+ | Planted defect categories caught | `6 / 6` (`21` defects: `10 P1`, `11 P2`) | `6 / 6` (`16` defects: `8 P1`, `8 P2`) | Equal recall; Arm 1 found 5 more defects |
88
+ | Total sessions (parent + child workers) | `4` (`1` coordinator + `3` workers) | `9` (`1` parent + `8` subagents) | `2.25x` fewer sessions (`3` vs `8` children) |
89
+ | Wall-clock runtime | `6m 30s` (`390s`) | `22m 41s` (`1,361s`) | `3.49x` faster (`71.3%` less time) |
90
+ | Combined transcript steps | `178` (`88` parent + `90` workers) | `655` (`19` parent + `636` subagents) | `3.68x` fewer steps (`72.8%` reduction) |
91
+ | Combined assistant turns | `49` (`25` parent + `24` workers) | `319` (`8` parent + `311` subagents) | `6.51x` fewer turns (`84.6%` reduction) |
92
+ | Combined tool calls | `117` (`57` parent + `60` workers) | `311` (`6` parent + `305` subagents) | `2.66x` fewer tool calls (`62.4%` reduction) |
93
+ | Combined raw transcript size | `336.8 KB` (`183.6 KB` workers) | `801.8 KB` (`773.7 KB` subagents) | `58.0%` smaller overall (`76.3%` child-only) |
94
+ | Parent inbox wakeup bytes | `1,629 B` (`5` notices; `647 B` handoffs) | `18,322 B` (`4` wakeups) | `91.1%` fewer parent wakeup bytes (`11.2x`) |
95
+ | Independent proof & quality outcome | SHA-256 bundle (`ad876bd4...` Digest) + `audit-ab-audit` `CLEAN` | `Victory Auditor` approved Sentinel-fabricated snippets; caught by parent (`Steps 11–17`) | Cryptographic proof vs. unverified QA |
96
+ | Scratch files left after run | `4` proof/report files (`72.3 KB`, gitignored) | `33` files (`183.7 KB` in `.agents/teamwork/`) | `29` fewer scratch files |
97
+
98
+ See [Benchmarks](docs/benchmarks.md) for the per-agent breakdown of the November 15 matched-`High` retest and the historical November 8 Phase 1 baseline (`6m 15s` vs. `26m 19s`, `148.2 KB` vs. `817.4 KB` subagent-only).
136
99
 
137
- | Metric | Agy-Squad + Agy-Context-Saver 1.8.1 (`rtk 0.51.0`) | Antigravity `/teamwork` | Difference |
138
- | --- | --- | --- | --- |
139
- | Planted defect categories caught | `6 / 6` | `6 / 6` | Parity |
140
- | Subagents spawned | `3` | `8` | `5` fewer |
141
- | Wall-clock runtime | `6m 15s` (`375s`) | `26m 19s` (`1,579s`) | `4.2x` faster |
142
- | Transcript steps | `64` | `645` | `90.1%` fewer steps |
143
- | Raw transcript size | `148.2 KB` | `817.4 KB` | `81.9%` smaller |
144
- | Parent context wakeup bytes | `249 B` | `4,940 B` | `95.0%` fewer bytes |
145
- | Git-untracked files left (`git status`) | `0` | `41` | `41` fewer |
100
+ ## Documentation
146
101
 
147
- In the Agy-Squad arm, the squad lead inspected cited source files directly and verified SHA-256 input/report digests in `20 ms` (`handoffs.mjs verify`) before pruning loose specialist reports under the 3-scope teardown policy.
102
+ Full documentation is hosted at [paragon-ux.github.io/Agy-Squad](https://paragon-ux.github.io/Agy-Squad/) (built with Zensical) and in [`docs/`](docs/):
148
103
 
149
- See [docs/benchmarks.md](docs/benchmarks.md) for the canonical comparison table, unit definitions, prompts, and live reproduction protocol. Run `node benchmarks/teamwork-vs-squad/verify-benchmark.mjs` to execute the offline fixture, manifest compilation, and telemetry ledger check.
104
+ - [Overview](docs/index.md)
105
+ - [Getting started](docs/getting-started.md)
106
+ - [Core workflows: orchestrate, self-diagnose, review, and tear down](docs/core-workflows.md)
107
+ - [Roles, skill workflows, and runtime closure checklist](docs/roles-and-workflows.md)
108
+ - [Concepts and terminology](docs/concepts.md)
109
+ - [Manifest and tool contract](docs/tools.md)
110
+ - [Stopping agents and child processes](docs/stopping-agents.md)
111
+ - [Benchmarks vs. Antigravity Teamwork](docs/benchmarks.md)
112
+ - [WSL CLI verification and live trial](docs/plans/wsl-cli-verification.md)
113
+ - [Architecture decision ledger](docs/plans/LEDGER.md)
114
+ - [Native enforcement ledger](docs/plans/native-enforcement-ledger.md)
150
115
 
151
116
  ## Development
152
117
 
153
- Run `npm test` with Node.js `>= 22`. The suite tests the MCP server, `dispatch_squad` compiler, `squad_planner` planner, handoff bundler, `/self-diagnosis` profiler, and offline benchmark verifier without launching live subagents.
118
+ Run `node scripts/release-preflight.mjs`, `npm test`, and `npm run verify:package` with Node.js `>= 22`. Runs all 10 test suites (`tests/test-verdicts-and-pivot.mjs`, `tests/test-dispatch.mjs`, `tests/test-planner.mjs`, `tests/test-handoffs.mjs`, `tests/test-run.mjs`, `tests/test-server.mjs`, `tests/test-comparative-dossier.mjs`, `tests/test-lifecycle-installer.mjs`, `benchmarks/teamwork-vs-squad/verify-benchmark.mjs`, and `tests/verify-package.mjs`) in isolated temporary directories without mutating your live `~/.gemini` configuration.
154
119
 
155
120
  ## License
156
121
 
157
- MIT
122
+ MIT License. Copyright (c) 2026 paragon-ux.
package/bin/agy-squad.mjs CHANGED
@@ -12,9 +12,9 @@ import { printStatus, runCheckOnly, runInstall, runRestoreBackups, runUninstall
12
12
  const __dirname = path.dirname(fileURLToPath(import.meta.url));
13
13
  const pkgVersion = (() => {
14
14
  try {
15
- return JSON.parse(fs.readFileSync(path.resolve(__dirname, "../package.json"), "utf8")).version || "0.1.0";
15
+ return JSON.parse(fs.readFileSync(path.resolve(__dirname, "../package.json"), "utf8")).version || "0.1.1";
16
16
  } catch {
17
- return "0.1.0";
17
+ return "0.1.1";
18
18
  }
19
19
  })();
20
20
 
package/docs/benchmarks.md CHANGED
@@ -1,8 +1,60 @@
1
1
  # Benchmarks: Agy-Squad vs. Antigravity Teamwork
2
2
 
3
- On November 8, 2026, we ran a recorded sequential comparison between **Agy-Squad + Agy-Context-Saver 1.8.1 (`rtk 0.51.0`)** and **Antigravity `/teamwork` (`[skip-interview]` control)** on a two-domain event-relay fixture containing six planted `P1` and `P2` defects ([`benchmarks/teamwork-vs-squad/`](https://github.com/paragon-ux/Agy-Squad/tree/main/benchmarks/teamwork-vs-squad)).
3
+ This page records two head-to-head sequential evaluations on the two-domain event-relay fixture (`6` planted `P1`/`P2` defects, [`benchmarks/teamwork-vs-squad/`](https://github.com/paragon-ux/Agy-Squad/tree/main/benchmarks/teamwork-vs-squad)):
4
4
 
5
- ## Canonical recorded comparison (November 8, 2026)
5
+ 1. **November 15, 2026 Matched-`High` Retest (current Phase 2 path):** Native `/swarm` + `Agy-Context-Saver 2.0.1` (`rtk 1.1.0`) vs. Antigravity `/teamwork [skip-interview]`, pinned to `Gemini 4 Hydrogen (High)` on all parents and child workers.
6
+ 2. **November 8, 2026 Historical Baseline (retired Phase 1 path):** `agy-squad run --launch` + `Agy-Context-Saver 1.8.1` (`rtk 0.51.0`) vs. Antigravity `/teamwork [skip-interview]`.
7
+
8
+ ## Matched-`High` Retest (November 15, 2026 — Native `/swarm` + `Agy-Context-Saver 2.0.1`)
9
+
10
+ Both arms caught all six planted defect categories (`6 / 6`). Counting **all sessions end-to-end (`parent + workers`)**, Agy-Squad finished in **`6m 30s` (`390s`) vs. `22m 41s` (`1,361s`, `3.49x` faster)**, used **`4` vs. `9` total sessions** (`3` vs. `8` spawned child workers), executed **`178` vs. `655` transcript steps** (`72.8%` reduction; `90` vs. `636` child-only steps), generated **`336.8 KB` vs. `801.8 KB` of raw transcripts** (`58.0%` smaller overall; `183.6 KB` vs. `773.7 KB`, `76.3%` smaller child-only), and sent **`1,629 B` vs. `18,322 B` into the parent inbox** (`91.1%` fewer wakeup bytes).
11
+
12
+ | Metric | Agy-Squad `/swarm` (`ab_audit`, flat) | Antigravity `/teamwork` (`[skip-interview]`) | Recorded difference |
13
+ | --- | ---: | ---: | --- |
14
+ | **Model & effort tier (all parents + children)** | `Gemini 4 Hydrogen (High)` | `Gemini 4 Hydrogen (High)` | Exact matched-effort parity |
15
+ | **Planted defect categories caught** | **`6 / 6`** (`21` defects: `10 P1`, `11 P2`) | **`6 / 6`** (`16` defects: `8 P1`, `8 P2`) | Equal recall; Arm 1 found 5 more unique defects |
16
+ | **Active wall-clock time** | **`6m 30s` (`390s`)** (`385s` internal clock) | **`22m 41s` (`1,361s`)** | **`3.49x` faster (`71.3%` less time)** |
17
+ | **Time to first specialist delegation** | **`47s`** (`42s` from initial `date`) | **`244s`** (Sentinel `9s`, Orchestrator `127s`) | **`5.19x` faster first domain worker** |
18
+ | **Total sessions (`parent + child workers`)** | **`4`** (`1` coordinator + `3` `/swarm` workers) | **`9`** (`1` parent + `8` subagents) | **`2.25x` fewer sessions (`3` vs. `8` children)** |
19
+ | **Combined transcript steps (end-to-end)** | **`178`** (`88` parent + `90` workers) | **`655`** (`19` parent + `636` subagents) | **`3.68x` fewer steps (`72.8%` reduction)** |
20
+ | **Combined assistant LLM turns** | **`49`** (`25` parent + `24` workers) | **`319`** (`8` parent + `311` subagents) | **`6.51x` fewer turns (`84.6%` reduction)** |
21
+ | **Combined tool calls (errors)** | **`117`** (`57` + `60`; `5` errors) | **`311`** (`6` + `305`; `16` errors) | **`2.66x` fewer tool calls (`62.4%` reduction)** |
22
+ | **Combined raw transcript size (end-to-end)** | **`336.8 KB`** (`153.2 KB` + `183.6 KB`) | **`801.8 KB`** (`28.2 KB` + `773.7 KB`) | **`2.38x` smaller (`58.0%` byte reduction)** |
23
+ | **Child-only raw transcript size** | **`183.6 KB`** (`3` workers) | **`773.7 KB`** (`8` subagents) | **`4.21x` smaller (`76.3%` byte reduction)** |
24
+ | **Parent wakeups and injected bytes** | **`5` notices (`1,629 B`; `647 B` handoffs)** | **`4` wakeups (`18,322 B` incl. `15,839 B` Step 10 CLI output)** | **`11.2x` fewer parent bytes (`91.1%` reduction)** |
25
+ | **Busy-polling violations** | **`0`** | **`0`** | Both compliant |
26
+ | **Quality & integrity verification** | Bundle SHA-256 `4c28db49...`, Digest `ad876bd4...`, `audit-ab-audit` `CLEAN` | `Victory Auditor` approved Sentinel-fabricated snippets/symbols; caught by parent (`Steps 11–17`) | Cryptographic proof + trace audit vs. missed fabrication |
27
+ | **Scratch artifacts retained** | **`4` files (`72,320 B`, specialists pruned, gitignored)** | **`33` files (`183,670 B` in `.agents/teamwork/`)** | **`29` fewer files (`60.6%` fewer bytes)** |
28
+
29
+ ### Matched-`High` per-agent telemetry (November 15, 2026)
30
+
31
+ #### Arm 1: Agy-Squad native `/swarm` (`4` sessions total, `336,829 B` / `336.8 KB`)
32
+
33
+ | Role | Conversation ID | Steps | Turns | Tool calls | Errors | Raw transcript |
34
+ | --- | --- | ---: | ---: | ---: | ---: | ---: |
35
+ | Parent Coordinator & Verification Lead | `5e4a24e1-0dd6-4028-b690-7a731772f2bf` | `88` | `25` | `57` | `1` | `153.2 KB` (`153,216 B`) |
36
+ | `ab-audit-gateway-securit` | `583dd7ae-d639-4bbd-91f4-f810f4426542` | `21` | `8` | `11` | `0` | `38.1 KB` (`38,121 B`) |
37
+ | `ab-audit-worker-retry-le` | `031e3bc7-7721-4427-a5e8-f531be0176ce` | `23` | `9` | `12` | `0` | `41.7 KB` (`41,722 B`) |
38
+ | `audit-ab-audit` (Integrity Auditor) | `15f815c6-7dec-480c-b7e8-3f6a71256f9c` | `46` | `7` | `37` | `4` | `103.8 KB` (`103,770 B`) |
39
+
40
+ > [!NOTE]
41
+ > **Friction investigated and fixed in `v0.1.1`:** During Arm 1 above, Parent Step 7 hit `1` `ENOENT` error when `mcp_agy-squad_squad_planner` ran with MCP server `cwd` outside the repository root (resolved via optional `workspace` parameter, `initialize` `rootUri` fallback, and actionable `ENOENT` diagnostics), and `audit-ab-audit` hit `4` pre-tool hook denials on unbounded `read_file` in Steps 5–8 before falling back to `rtk read` (resolved by adding bounded `<=150`-line / `rtk read` guidance in `mcp/briefs.mjs`).
42
+
43
+ #### Arm 2: Antigravity `/teamwork` control (`9` sessions total, `801,839 B` / `801.8 KB`)
44
+
45
+ | Role | Conversation ID | Steps | Turns | Tool calls | Errors | Raw transcript |
46
+ | --- | --- | ---: | ---: | ---: | ---: | ---: |
47
+ | Parent Session | `c657d27b-892e-487a-b461-1ebc5484e860` | `19` | `8` | `6` | `0` | `28.2 KB` (`28,177 B`) |
48
+ | Sentinel | `0d03a8bb-a10b-48b1-8996-283e6ed47611` | `97` | `46` | `44` | `0` | `94.3 KB` (`94,290 B`) |
49
+ | Orchestrator | `1912c806-59e3-4df8-bc10-ee70b955bca9` | `102` | `51` | `45` | `6` | `136.0 KB` (`136,047 B`) |
50
+ | Security Explorer | `08eb6ae7-b109-43cb-95ac-b6dd0e6a8942` | `65` | `29` | `35` | `4` | `85.6 KB` (`85,609 B`) |
51
+ | Reliability Explorer | `9aa12cb7-aac5-43b1-9b83-7529002f8b5a` | `74` | `37` | `36` | `1` | `88.8 KB` (`88,837 B`) |
52
+ | Contract Explorer | `3d6d877a-ebfb-4e7c-b152-39874f9ee5ca` | `81` | `40` | `40` | `2` | `104.5 KB` (`104,516 B`) |
53
+ | Quality Reviewer | `b6a2378b-cce9-412b-a8b7-8fb117ba3f07` | `62` | `31` | `30` | `2` | `87.1 KB` (`87,072 B`) |
54
+ | Forensic Auditor | `65936336-840e-46e5-8556-1fb6f5bd3214` | `76` | `38` | `37` | `1` | `77.2 KB` (`77,155 B`) |
55
+ | Victory Auditor | `0c070377-6818-41cc-ad82-b53fa4ae15bf` | `79` | `39` | `38` | `0` | `100.1 KB` (`100,136 B`) |
56
+
57
+ ## Historical baseline comparison (November 8, 2026)
6
58
 
7
59
  Both arms caught all six planted defect categories (`6 / 6`). Agy-Squad + Agy-Context-Saver finished in **`6m 15s` (`375s`) vs. `26m 19s` (`1,579s`, `4.2x` faster)**, used **`64` vs. `645` transcript steps** (`10.1x` fewer steps), generated **`148.2 KB` vs. `817.4 KB` of raw transcripts** (`81.9%` smaller), and sent **`249 B` vs. `4,940 B` into the parent conversation** (`95.0%` fewer parent wakeup bytes).
8
60
 
@@ -30,11 +82,11 @@ Both arms caught all six planted defect categories (`6 / 6`). Agy-Squad + Agy-Co
30
82
 
31
83
  ## Recorded session metrics and unit definitions
32
84
 
33
- All session counts in `benchmarks/teamwork-vs-squad/results/telemetry.json` come from `agy-context-saver`'s `check_context_health` tool:
85
+ Session counts in [`benchmarks/teamwork-vs-squad/results/telemetry.json`](https://github.com/paragon-ux/Agy-Squad/blob/main/benchmarks/teamwork-vs-squad/results/telemetry.json) come from `agy-context transcript health --json` in the November 15, 2026 matched-`High` retest (`Agy-Context-Saver 2.0.1`) and from `check_context_health` in the November 8, 2026 baseline (`Agy-Context-Saver 1.8.1`):
34
86
 
35
- - **Transcript steps, turns, and tool calls:** Operational counts from `.system_generated/logs/transcript.jsonl` across every spawned subagent in each arm.
36
- - **Raw transcript size (`148.2 KB` vs. `817.4 KB`) and parent wakeup bytes (`249 B` vs. `4,940 B`):** Measured in decimal kilobytes (`1,000 B`) and UTF-8 message bytes. These measure transcript and parent-inbox volume, not billed API token charges.
37
- - **Finding adjudication (`6 / 6` categories vs. `18` and `25` total items):** Both arms caught all six planted defect categories (`N1` through `N6`). Total item counts (`18` in Agy-Squad vs. `25` in Teamwork) include sub-defects split out from `N6` (corrupt JSON reset, non-atomic write, off-by-one retry) plus secondary input-validation observations.
87
+ - **Transcript steps, turns, and tool calls:** Operational counts from `.system_generated/logs/transcript.jsonl`. The November 15 retest records both child-worker totals (`workersSummary` / `childrenSummary`) and full end-to-end totals including parent coordination (`allSessionsSummary`), whereas the November 8 Phase 1 baseline recorded child-subagent totals (`3` vs. `8`).
88
+ - **Raw transcript size and parent wakeup bytes:** Measured in decimal kilobytes (`1,000 B`) and UTF-8 message bytes (`336.8 KB` vs. `801.8 KB` end-to-end and `1,629 B` vs. `18,322 B` in November 15; `148.2 KB` vs. `817.4 KB` child-only and `249 B` vs. `4,940 B` in November 8). These measure transcript and parent-inbox volume, not billed API token charges.
89
+ - **Finding adjudication (`6 / 6` categories across all runs):** Both arms caught all six planted defect categories (`N1` through `N6`). Total reported item counts (`21` vs. `16` in November 15; `18` vs. `25` in November 8) depend on how `N6` sub-defects and secondary input-validation observations are itemized.
38
90
 
39
91
  ## Benchmark fixture and six planted bugs
40
92
 
@@ -49,7 +101,7 @@ The fixture in `benchmarks/teamwork-vs-squad/fixture/` models a two-stage webhoo
49
101
  | `N5` | `P1` | `worker/worker.mjs:12` vs. `config.json:3` | Worker compares millisecond timestamp delta `Math.abs(nowMs - event.timestampMs)` against `config.maxSkewSeconds` (`300`) without `* 1000`, shrinking the 5-minute window to `300 ms` | Yes (`SYN-05`) | Yes (`BUG-01`) |
50
102
  | `N6` | `P1` | `worker/ledger.mjs:9-21` & `worker/worker.mjs:16-28` | `readLedger` resets corrupt JSON to `[]` (wiping prior entries on non-atomic `writeFileSync`) and `attempt >= maxRetries` on 1-based `attempt` performs `0` retries when `maxRetries = 1` | Yes (`SYN-06`) | Yes (`REL-02` / `BUG-02`) |
51
103
 
52
- ## Per-agent telemetry breakdown
104
+ ## Per-agent telemetry breakdown (November 8, 2026 Phase 1 baseline)
53
105
 
54
106
  ### Arm 1: Agy-Squad + Agy-Context-Saver (`3` subagents, `148.2 KB` total)
55
107
 
package/docs/core-workflows.md CHANGED
@@ -1,6 +1,6 @@
1
1
  # Core workflows: orchestrate, self-diagnose, review, and tear down
2
2
 
3
- Agy-Squad runs four connected workflows across **Antigravity Desktop (IDE)** and the **Antigravity CLI (`agy`)** on **Linux and WSL2 (macOS support is best-effort and untested)**:
3
+ Agy-Squad runs four connected workflows across **Antigravity Desktop (IDE)** and the **Antigravity CLI (`agy`)** on **Linux and WSL2**:
4
4
 
5
5
  1. **Task-proportional orchestration** (`squad_planner` / `agy-squad plan`, `dispatch_squad` / `agy-squad dispatch`, and `node bin/agy-squad.mjs` backed by `mcp/handoffs.mjs`)
6
6
  2. **Recursive harness self-diagnosis** ([`.agents/skills/self-diagnosis/SKILL.md`](https://github.com/paragon-ux/Agy-Squad/blob/main/.agents/skills/self-diagnosis/SKILL.md))
package/docs/getting-started.md CHANGED
@@ -1,12 +1,12 @@
1
1
  # Getting started (Desktop IDE & CLI on POSIX / WSL2)
2
2
 
3
- Agy-Squad is a subagent planning, self-diagnosis, and review harness for **Antigravity Desktop (IDE)** and the **Antigravity CLI (`agy`)** across **Linux and WSL2 (macOS support is best-effort and untested)**. It pairs [`swarm-governance`](roles-and-workflows.md#swarm-governance) (Execution Modifier) with Task-Spec skills ([`self-diagnosis`](core-workflows.md#2-diagnose-and-optimize-the-harness-self-diagnosis), [`pr-review`](core-workflows.md#3-review-pull-requests-to-the-v140-standard), and [`full-review`](roles-and-workflows.md#full-review)) using `<= 500` byte `Pointer-Over-Wire` receipts, file-backed evidence slots, and SHA-256 bundle checks (`v1.4.0`).
3
+ Agy-Squad is a subagent planning, self-diagnosis, and review harness for **Antigravity Desktop (IDE)** and the **Antigravity CLI (`agy`)** across **Linux and WSL2**. It pairs [`swarm-governance`](roles-and-workflows.md#swarm-governance) (Execution Modifier) with Task-Spec skills ([`self-diagnosis`](core-workflows.md#2-diagnose-and-optimize-the-harness-self-diagnosis), [`pr-review`](core-workflows.md#3-review-pull-requests-to-the-v140-standard), and [`full-review`](roles-and-workflows.md#full-review)) using `<= 500` byte `Pointer-Over-Wire` receipts, file-backed evidence slots, and SHA-256 bundle checks (`v1.4.0`).
4
4
 
5
5
  ## Host and companion compatibility matrix
6
6
 
7
7
  | Layer | Requirement | Check command |
8
8
  | --- | --- | --- |
9
- | Operating system | Linux or WSL2 on Windows (primary verified targets; macOS is best-effort and untested) | `uname -a` |
9
+ | Operating system | Linux or WSL2 on Windows | `uname -a` |
10
10
  | Node.js | `>= 22.0.0` | `node --version` |
11
11
  | Antigravity surface | Antigravity Desktop (IDE) with `/swarm` and/or `invoke_subagent` | Open Desktop IDE |
12
12
  | Agy-Context-Saver | `>= 2.0.1` (`PreToolUse` hook + `agy-context` CLI active) for live governed cohorts | `agy-context status` (or `agy-context setup`) |
package/docs/index.md CHANGED
@@ -1,6 +1,6 @@
1
1
  # Agy-Squad
2
2
 
3
- Agy-Squad is a subagent planning, self-diagnosis, and review harness for **Antigravity Desktop (IDE)** and the **Antigravity CLI (`agy`)** on **Linux and WSL2 (macOS support is best-effort and untested)**.
3
+ Agy-Squad is a subagent planning, self-diagnosis, and review harness for **Antigravity Desktop (IDE)** and the **Antigravity CLI (`agy`)** on **Linux and WSL2**.
4
4
 
5
5
  `squad_planner` and `dispatch_squad` assign scoped tasks to specialist or lead cohorts, cap `Pointer-Over-Wire` receipts at `500` UTF-8 bytes, and pair with `node bin/agy-squad.mjs` (`mcp/handoffs.mjs`) to verify SHA-256 input and report digests. During live Desktop IDE runs, `Agy-Context-Saver` (`2.0.1+`) rewrites shell commands through `rtk >= 1.1.0` (`rtk hook antigravity`), enforces 5-tier transcript access governance, and applies Reactive Wakeup backoff.
6
6
 
@@ -14,7 +14,7 @@ Agy-Squad is a subagent planning, self-diagnosis, and review harness for **Antig
14
14
  - **[Concepts and terminology](concepts.md):** Definitions for stages (`all_l1` vs. `<squadId>` vs. `integrity_auditor` / `audit-<id>`), `squad_planner`, report slots, `Pointer-Over-Wire` receipts, SHA-256 bundles, Reactive Wakeup, `RTK`, and decision ledger codes (`L1` through `L43`).
15
15
  - **[Manifest and tool contract](tools.md):** Schema and CLI flags for `squad_planner` (`plan`), `dispatch_squad` (`dispatch`), `node bin/agy-squad.mjs`, and `mcp/handoffs.mjs` (library module).
16
16
  - **[Native enforcement ledger](plans/native-enforcement-ledger.md):** Host hook boundary (`PreToolUse`, `StopHook`), `/swarm` channel limits, and Spike status.
17
- - **[Benchmarks](benchmarks.md):** Canonical metric table, unit definitions, and reproduction guide for the November 8, 2026 audit against Antigravity `/teamwork`.
17
+ - **[Benchmarks](benchmarks.md):** Canonical metric tables, per-agent telemetry, and reproduction guide for the November 15, 2026 matched-`High` `/swarm` retest and November 8, 2026 Phase 1 baseline against Antigravity `/teamwork`.
18
18
 
19
19
  ## Core capabilities
20
20
 
@@ -38,10 +38,10 @@ Run `agy-context setup` once on a new host so the `rtk >= 1.1.0` binary and `agy
38
38
 
39
39
  ## Recorded comparison vs. Antigravity `/teamwork`
40
40
 
41
- On November 8, 2026, a sequential audit compared both approaches on the six-defect event-relay fixture ([`benchmarks.md`](benchmarks.md)). Both caught all six planted defect categories (`6 / 6`).
41
+ On November 15, 2026, a matched-effort sequential retest (`Gemini 4 Hydrogen (High)` on all parents and child workers) compared native `/swarm` (`Agy-Squad` + `Agy-Context-Saver 2.0.1`, `rtk 1.1.0`) against `/teamwork [skip-interview]` on the six-defect event-relay fixture ([`benchmarks.md`](benchmarks.md)). Both arms caught all six planted defect categories (`6 / 6`):
42
42
 
43
- - **Runtime and steps:** **Agy-Squad + Agy-Context-Saver (`1.8.1`, `rtk 0.51.0`)** finished in **`6m 15s` (`375s`) vs. `26m 19s` (`1,579s`, `4.2x` faster)** and executed **`64` vs. `645` transcript steps** (`90.1%` fewer steps).
44
- - **Context and transcript size:** Raw transcripts totaled **`148.2 KB` vs. `817.4 KB`** (`81.9%` smaller), and parent context wakeup payloads totaled **`249 B` vs. `4,940 B`** (`95.0%` fewer wakeup bytes).
45
- - **Verification and cleanup:** The squad lead verified cited fixture lines directly and ran a 20 ms SHA-256 bundle digest check (`handoffs.mjs verify` during the Phase 1 trial; now `node bin/agy-squad.mjs verify`) before pruning loose reports under the 3-scope teardown policy.
43
+ - **Runtime and steps (end-to-end parent + all workers):** **Agy-Squad `/swarm`** finished in **`6m 30s` (`390s`) vs. `22m 41s` (`1,361s`, `3.49x` faster)** and executed **`178` vs. `655` combined transcript steps** (`72.8%` fewer steps; `90` vs. `636` child-only steps).
44
+ - **Context, transcript size, and wakeups:** Combined raw transcripts totaled **`336.8 KB` (`183.6 KB` child-only) vs. `801.8 KB` (`773.7 KB` child-only)** (`58.0%` smaller overall, `76.3%` smaller child-only), and parent wakeup payloads totaled **`1,629 B` vs. `18,322 B`** (`91.1%` fewer parent bytes).
45
+ - **Verification and evidence quality:** Agy-Squad found **`21` unique defects (`10 P1`, `11 P2`)** backed by SHA-256 bundle `4c28db49...` and an independent `audit-ab-audit` `CLEAN` verdict, whereas `/teamwork` found **`16` defects (`8 P1`, `8 P2`)** and its `Victory Auditor` approved Sentinel-fabricated snippets that the parent session had to catch and fix in Steps 11–17.
46
46
 
47
- See [Benchmarks](benchmarks.md) for the full metric table, unit definitions, prompts, and offline vs. live reproduction steps.
47
+ See [Benchmarks](benchmarks.md) for the complete November 15 matched-`High` per-agent table, the historical November 8, 2026 Phase 1 baseline (`6m 15s` vs. `26m 19s`, `148.2 KB` vs. `817.4 KB` subagent-only), unit definitions, and reproduction steps.