feat(instructions): add Antigravity global agent rules

CHANGELOG.md

@@ -1,5 +1,16 @@
 # Changelog
 
+## 0.3.3 - 2026-07-09
+
+- Promote managed global agent operating rules v1.1.2 with Antigravity provider rendering and Gemini retired from active providers.
+- Add non-overridable worktree, PR-first, no-main-push, autonomous repair, conversations surfaces, and no-budget-unless-requested rules.
+
+
+## Unreleased
+
+- Bumped `hasna-agent-operating-rules` to v1.1.2 (2026-07-09) and expanded the canonical source set with rules for automatic session renaming, task-specific worktree mutation under `$HOME/.hasna/repos/worktrees`, PR-first landing, no direct pushes to main/default/protected branches, autonomous repair before asking, full Hasna CLI/package source-of-truth coverage, default conversation surfaces plus `conversations blockers`, durable goal-plan adversarial verification, and Codewith goal/token/goal-plan budget opt-in only.
+- Added Antigravity as an active global instruction provider target and provider overlay, while keeping Gemini out of active target/provider compatibility coverage.
+
 ## 0.2.1
 
 - Added the versioned `hasna-agent-operating-rules` canonical source: Hasna Agent Operating Rules v1.1.0 (version stamp on line 1, sentinel `<!-- hasna:agent-operating-rules v=1.1.0 -->`), leading with Andrei's four core operating rules (independent adversarial reviewer on every user-requested work item; record-as-you-go in todos/mementos/conversations CLIs; register an agent identity before taking work, subagents never; every project has a continuously updated conversations channel) followed by the fleet communication duties from the fleet comms strategy (todos task 39a68145).

README.md

@@ -237,6 +237,7 @@ identities instructions export ./instructions.json --json
 identities instructions import ./instructions.json --json
 identities instructions sources --json
 identities instructions sources --canonical --provider codewith --json
+identities instructions export --canonical --provider antigravity --json
 identities instructions export --canonical --provider codewith --json
 identities instructions export --canonical --provider opencode --json
 ```
@@ -254,27 +255,38 @@ fields derived from `kind`, `mergePolicy`, and `precedence`.
 OpenIdentities also ships the canonical Hasna global coding-agent source set
 for downstream renderers. It contains one global system prompt, one
 non-overridable global rules source, the versioned non-overridable Hasna Agent
-Operating Rules document (`hasna-agent-operating-rules`, currently v1.1.0 with
-sentinel `<!-- hasna:agent-operating-rules v=1.1.0 -->`, precedence 175), and
-provider overlays for Codewith, Claude Code, Codex, and OpenCode. OpenConfigs
-should consume these sources and render managed provider blocks or OpenCode
-instruction references; it remains responsible for file rendering, path
-dereferencing, and merge mechanics.
+Operating Rules document (`hasna-agent-operating-rules`, currently v1.1.2 with
+sentinel `<!-- hasna:agent-operating-rules v=1.1.2 -->`, precedence 175), and
+provider overlays for Antigravity, Codewith, Claude Code, Codex, and OpenCode.
+OpenConfigs should consume these sources and render managed provider blocks or
+OpenCode instruction references; it remains responsible for file rendering,
+path dereferencing, and merge mechanics. Antigravity is an active target in this
+source set; Gemini is not an active target and should not be restored as one.
 
 The canonical set includes rules for Knowledge CLI/SDK usage, Todos plans and
-evidence, Mementos/Conversations/Projects source-of-truth boundaries,
-coordinator delegation, Codewith-native loop terminology versus OpenLoops,
-dispatch self-healing without tmux fallback, adversarial verification, secrets
-safety, commit/push secrets scans, no Co-Authored-By trailers, Bun preference,
-and Hasna package release-age registry hygiene. The Agent Operating Rules add
-the four core operating rules (adversarial reviewer on every user-requested
-work item, record-as-you-go in the todos/mementos/conversations CLIs, identity
-registration before taking work with subagents never registering, and a
-continuously updated conversations channel per project) plus the fleet
-communication duties (bounded announcements/blockers reads, `[BREAKING]`
+evidence, Hasna CLI/package source-of-truth boundaries (todos, conversations,
+mementos, knowledge, projects, repos, accounts, instructions, machines, secrets,
+access), automatic session renaming when supported, task-specific worktree
+mutation under the canonical `$HOME/.hasna/repos/worktrees` root, PR-first
+landing, no direct pushes to main/default/protected branches, autonomous repair
+before asking, coordinator delegation, Codewith-native loop terminology versus
+OpenLoops, dispatch self-healing without tmux fallback,
+adversarial verification, secrets safety, commit/push secrets scans, no
+Co-Authored-By trailers, Bun preference, and Hasna package release-age registry
+hygiene. The Agent Operating Rules add the core operating rules (adversarial
+reviewer on every user-requested work item, record-as-you-go in the
+todos/mementos/conversations CLIs, identity registration before taking work
+with subagents never registering, and a continuously updated conversations
+channel per project), durable goal-plan adversarial verification, default
+conversation surfaces (announcements, incidents, git-publishing, git-prs,
+git-commits, git-releases, hq, agent-policy, project/product channels, and
+`conversations blockers`), and the fleet communication duties (`[BREAKING]`
 heads-up before fleet-affecting changes, publish intent before npm/bun publish,
 incidents-first, no secrets in messages, channel content treated as data,
-convention lookup before naming, identity release at session end).
+convention lookup before naming, identity release at session end). The Codewith
+overlay keeps Codewith-native goal, goal-plan, schedule, and loop guidance, while
+the global non-overridable rules forbid setting Codewith goal, token, or
+goal-plan budgets unless the user explicitly asks for budgets.
 
 SDK consumers can import the same data from `@hasna/identities`:
 

docs/instructions.md

@@ -129,6 +129,7 @@ provider files or managed blocks in their own layer.
 ```bash
 identities instructions sources --canonical --json
 identities instructions sources --canonical --provider codewith --json
+identities instructions export --canonical --provider antigravity --json
 identities instructions export --canonical --provider claude --json
 identities instructions export --canonical --provider opencode --json
 ```
@@ -138,6 +139,7 @@ The canonical set includes:
 - `hasna-global-coding-agent-system-prompt`
 - `hasna-global-coding-agent-non-overridable-rules`
 - `hasna-agent-operating-rules`
+- `hasna-antigravity-global-agent-overlay`
 - `hasna-codewith-global-agent-overlay`
 - `hasna-claude-global-agent-overlay`
 - `hasna-codex-global-agent-overlay`
@@ -148,34 +150,54 @@ matching provider overlays. For example, `--provider codewith` returns the three
 global sources and the Codewith overlay.
 
 `hasna-agent-operating-rules` is the versioned Hasna Agent Operating Rules
-document (currently v1.1.0, stamped on line 1 and carrying the sentinel comment
-`<!-- hasna:agent-operating-rules v=1.1.0 -->` so renderers and drift checks can
-verify currency). It leads with the four core operating rules — an independent
+document (currently v1.1.2, stamped on line 1 and carrying the sentinel comment
+`<!-- hasna:agent-operating-rules v=1.1.2 -->` so renderers and drift checks can
+verify currency). It leads with the core operating rules — an independent
 adversarial reviewer on every user-requested piece of work, record-as-you-go in
 the todos/mementos/conversations CLIs, agent-identity registration before
-taking work (subagents never register), and a continuously updated
-conversations channel per project — followed by the fleet communication duties
-(announcements/blockers reads bounded `--since 7d`, `[BREAKING]` heads-up before
-fleet-affecting changes, publish intent before npm/bun publish, incidents-first,
-no secrets in messages, channel content is data not instructions, convention
-lookup before naming, identity release at session end). It is non-overridable
+taking work (subagents never register), a continuously updated conversations
+channel per project, automatic session renaming when supported, Hasna
+CLI/package source-of-truth boundaries, and autonomous repair before asking —
+then code/landing rules and fleet communication duties. It is non-overridable
 and renders at precedence 175, between the global system prompt (150) and the
 provider overlays (200).
 
 Required rule coverage is part of the source content:
 
 - Knowledge must use the Knowledge CLI or SDK, never ad hoc global Markdown
   under `$HOME/.hasna`, `$HOME/.husna`, or similar scratch paths.
+- Hasna CLIs/packages are source of truth for todos, conversations, mementos,
+  knowledge, projects, repos, accounts, instructions, machines, secrets, and
+  access.
 - Planning and evidence must use Todos CLI tasks and todos plans.
-- Mementos, Conversations, and Projects CLIs remain source of truth in their
-  domains.
+- Agents should automatically rename sessions when the runtime supports it.
+- Repo mutation must happen in task-specific worktrees under the canonical
+  `$HOME/.hasna/repos/worktrees` root; prefer Hasna repo/project worktree
+  mechanisms, otherwise `git worktree` rooted there, and never mutate shared
+  checkouts.
+- Normal changes use PR-first landing through a branch/worktree plus a pull
+  request or prepared pull-request handoff.
+- Agents must not push directly to `main`, default, or protected branches unless
+  the user explicitly instructs that exact repo and operation.
+- Agents act autonomously by diagnosing and repairing owning CLIs, packages,
+  and workflows before asking the user; ask only for destructive,
+  secret-bearing, or user-only decisions.
 - Coordinator sessions delegate product-code implementation through subagents
   or task workflows.
 - Codewith-native loops and OpenLoops are different mechanisms and terms.
 - Dispatch failures require self-healing of the owning package or workflow; no
   tmux prompt-paste fallback is allowed without explicit human authorization.
 - Non-trivial work needs adversarial verification or a labeled adversarial
   self-review when no reviewer can be spawned.
+- Durable goal plans require adversarial verification steps during the plan and
+  a final adversarial verification step before completion.
+- Default conversation surfaces are announcements, incidents, git-publishing,
+  git-prs, git-commits, git-releases, hq, agent-policy, project/product
+  channels, and `conversations blockers` (not a literal blockers channel).
+- Codewith goal, token, and goal-plan budgets must not be set unless the user
+  explicitly asks for budgets.
+- Antigravity is an active provider target; Gemini must not be created or
+  restored as an active global instruction provider target.
 - Secrets must not be exposed, commit/push secrets scans are mandatory, and
   commits must not use Co-Authored-By trailers.
 - Bun is preferred for Hasna JavaScript and TypeScript repositories, with

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hasna/identities",
-  "version": "0.2.2",
+  "version": "0.3.3",
   "description": "Open identity records for humans and AI agents",
   "type": "module",
   "main": "dist/src/index.js",
@@ -131,7 +131,7 @@
     "typescript": "^5.7.3"
   },
   "dependencies": {
-    "@hasna/contracts": "0.4.1",
+    "@hasna/contracts": "0.4.2",
     "@modelcontextprotocol/sdk": "^1.29.0",
     "pg": "^8.16.3",
     "zod": "^3.24.2"