[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CT4T54N17MTY55F562WKVP":3,"skill-same:01M7CT4T54N17MTY55F562WKVP":44},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M7CT4T54N17MTY55F562WKVP","antigravity-sdk-e2e-dev",null,"Spin up a live local Omnigent server and exercise the Antigravity (Gemini) SDK harness end-to-end — build antigravity agents, run real turns, smoke-test, and bug-bash. Load when developing, testing, or debugging the antigravity harness (omnigent\u002Finner\u002Fantigravity_executor.py, antigravity_harness.py, omnigent\u002Fonboarding\u002Fantigravity_auth.py) or its auth \u002F model \u002F tool-bridge behavior.","roland-id-au\u002Fomniagent","roland-id-au","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F2590653?v=4","01M7CT4DEBVK4R185WFWCWN4JD","Buck Chancey","01M7CT4DFGWX49CGE347EF4V6T",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent\u002Fblob\u002Fdeploy\u002Froland-id-au-omniagent\u002F.claude\u002Fskills\u002Fantigravity-sdk-e2e-dev\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: antigravity-sdk-e2e-dev\ndescription: Spin up a live local Omnigent server and exercise the Antigravity (Gemini) SDK harness end-to-end — build antigravity agents, run real turns, smoke-test, and bug-bash. Load when developing, testing, or debugging the antigravity harness (omnigent\u002Finner\u002Fantigravity_executor.py, antigravity_harness.py, omnigent\u002Fonboarding\u002Fantigravity_auth.py) or its auth \u002F model \u002F tool-bridge behavior.\n---\n\n# Antigravity SDK harness: end-to-end dev & testing\n\nThe `antigravity` harness drives Google's **Antigravity Python SDK**\n(`google-antigravity`, an in-process `Agent`\u002F`Conversation`) and bridges\nOmnigent's `sys_*` tools into the SDK as `custom_tools`. It is **Gemini-native**:\nit authenticates with a Gemini \u002F Antigravity API key (or Vertex AI) and has **no\nOpenAI-compatible gateway \u002F Databricks path**. This skill is the proven recipe\nfor running it **for real** against a live local server — not just the unit\ntests.\n\n> The harness runs as a **local runner** from your current checkout, so\n> `omni run \u003Cbundle> --server \u003Curl>` exercises exactly the code you're on.\n\n## Prerequisites (check these first)\n\n1. **You're on the branch you want to test.** The antigravity harness merged to\n   `main` (#194). Test on `main` unless validating a specific branch.\n2. **A Gemini API key is configured.** The SDK *requires* one (`AIza…`); there\n   is no login flow. Verify (booleans only — never print the key):\n   ```bash\n   .venv\u002Fbin\u002Fpython -c \"from omnigent.onboarding.antigravity_auth import antigravity_api_key_configured as c; import os; print('config:', c(), 'env:', bool(os.environ.get('GEMINI_API_KEY') or os.environ.get('ANTIGRAVITY_API_KEY')))\"\n   ```\n   If both are `False`, run `omni setup` → **Antigravity** and paste a key, or\n   `export GEMINI_API_KEY=AIza…`.\n3. **`google-antigravity` is installed** (the `antigravity` extra —\n   `pip install \"omnigent[antigravity]\"`):\n   `.venv\u002Fbin\u002Fpython -c \"import google.antigravity as a; print(a.__file__)\"`.\n4. **glibc ≥ ~2.36.** The SDK spawns a **native `localharness` binary** that\n   needs a recent glibc (`GLIBC_ABI_DT_RELR`). Check `ldd --version | head -1`.\n   On an older host the turn fails at setup with\n   `RuntimeError: … localharness: … version 'GLIBC_ABI_DT_RELR' not found`. Dev\n   workaround on a glibc-2.31 box: point the SDK at a loader-shim via\n   `ANTIGRAVITY_HARNESS_PATH=\u002Fpath\u002Fto\u002Fshim` that runs the *untouched* bundled\n   binary through a newer glibc's loader (see the auto-memory note\n   `antigravity-harness-glibc-native-binary.md`). The shim is dev-only — the\n   real fix is a glibc-≥2.36 host.\n5. **Network egress to the Gemini backend.** The native binary talks to\n   Google's API; a turn that hangs or fails to connect on a locked-down host is\n   usually an egress problem, not a harness bug.\n\n## Step 1 — start a local server\n\n```bash\ncd \u002Fpath\u002Fto\u002Fomnigent\n.venv\u002Fbin\u002Fomni server --background          # spawns a detached server on a free loopback port\n.venv\u002Fbin\u002Fomni server status         # prints the URL, e.g. http:\u002F\u002F127.0.0.1:6767\n```\n\nUse the **printed URL** below as `$SERVER`. (You can also run a foreground\nserver on a fixed port with `omnigent server --port 7777 --no-open`.)\n\n## Step 2 — build an antigravity agent bundle\n\nA spec with `spec_version` **must be a directory containing `config.yaml`** —\nnot a single `.yaml` file. Minimal antigravity agent (no `auth:` block → it\nresolves the key from the `antigravity:` config \u002F ambient env):\n\n```bash\nmkdir -p \u002Ftmp\u002Fagy-dev\ncat > \u002Ftmp\u002Fagy-dev\u002Fconfig.yaml \u003C\u003C'YAML'\nspec_version: 1\nname: agy-dev\ndescription: Antigravity SDK dev\u002Ftest agent.\nexecutor:\n  type: omnigent\n  config:\n    harness: antigravity\n    model: gemini-3.5-flash      # default; gemini-3-pro 404s on a plain AI-Studio key\nprompt: |\n  You are a terse test agent. Answer in as few words as possible.\nYAML\n```\n\nFor sub-agents, tools, guardrails\u002Fpolicies, copy the field shapes from\n`examples\u002Fpolly\u002Fconfig.yaml` and `examples\u002Fdebby\u002Fconfig.yaml`.\n\n## Step 3 — run a turn (and smoke-test)\n\n```bash\nSERVER=http:\u002F\u002F127.0.0.1:6767   # the URL from `omni server status`\ntimeout 280 .venv\u002Fbin\u002Fomni run \u002Ftmp\u002Fagy-dev \\\n  -p \"Reply with exactly the single word: PONG\" \\\n  --server \"$SERVER\" 2>&1\n```\n\nA healthy run prints connection lines then the assistant reply (`PONG`). If\nthat works, the full stack is good: Gemini key, glibc\u002Fnative binary, egress,\nstreaming, harness.\n\n- **Shell \u002F file tools:** add `--tools coding`.\n- **Specific model:** add `--model gemini-2.5-flash` (or another Gemini id).\n\n## Targeted scenarios\n\n| Goal | How |\n|------|-----|\n| Native tools (shell\u002Fedit\u002Fread) | `--tools coding`, prompt to create→read→edit a file and run a shell command; confirm it actually touches disk |\n| Bridged `sys_*` \u002F sub-agent dispatch | declare a sub-agent (`tools.agents`\u002F`spawn`), prompt the agent to delegate — exercises the `custom_tools` bridge + `PostToolCallHook` |\n| Model routing | run the same bundle with several `--model` Gemini ids; note which actually runs |\n| Vertex AI auth | set `executor.config.vertex: true` + `project`\u002F`location` and use GCP application-default creds instead of an API key |\n| Policy \u002F guardrail | add a guardrail that denies a keyword; confirm it blocks (see the **sharp edges** below — LLM-phase + tool-call enforcement was incomplete at merge) |\n| Per-session brain override | run a bundle agent (polly\u002Fdebby) and select `antigravity` as the brain harness (it's in `BRAIN_HARNESS_LABELS`) |\n| Concurrency \u002F leaks | fire several `omni run … &` at once; then `pgrep -af localharness` to check for orphaned native subprocesses |\n\n## Gotchas (these cost real time)\n\n1. **`config.yaml`'s `server:` defaults to a *remote* server.** Omitting\n   `--server` sends your turn to that remote deploy — which may be **stale** and\n   reject the antigravity harness with `executor.config.harness: must be one of\n   […], got 'antigravity'`. **Always pass `--server http:\u002F\u002F127.0.0.1:\u003Cport>`**\n   for local testing. (That allowlist is `omnigent\u002Fspec\u002F_omnigent_compat.py`; if\n   a *local* server rejects `antigravity`, it's running stale code — restart it\n   from your checkout.)\n2. **A spec with `spec_version` must be a directory + `config.yaml`**, never a\n   single `.yaml` file.\n3. **Antigravity needs a Gemini key** (no login). Resolution precedence: spec\n   `executor.auth` (api_key) > stored `antigravity:` config block (`omni setup`)\n   > ambient `GEMINI_API_KEY` \u002F `ANTIGRAVITY_API_KEY`. Vertex AI is opt-in via\n   `executor.config` `vertex`\u002F`project`\u002F`location`.\n4. **No OpenAI gateway \u002F Databricks.** The SDK has no `base_url`; a `databricks`\n   or generic-`provider` auth is **warned and ignored**, and the run falls back\n   to ambient Gemini creds. Don't expect `databricks-*` models to route through\n   the AI Gateway like claude-sdk\u002Fcodex\u002Fpi.\n5. **Model ids are Gemini ids.** Default `gemini-3.5-flash`. `gemini-3-pro`\n   **404s on a plain AI-Studio key** — use `gemini-2.5-flash` \u002F `gemini-3.5-flash`\n   unless your key has Pro access.\n6. **The native binary needs glibc ≥ ~2.36** (see Prereq 4). This is the most\n   common \"it won't even start\" cause; check it before assuming a harness bug.\n7. **Turns take ~10–60s** — always wrap in `timeout 280`.\n8. **Local-runner topology:** `omni run \u003Cbundle> --server \u003Curl>` runs the\n   harness from your **current checkout**; the server only holds state. The\n   managed `omni server --background` server runs from whatever venv launched it.\n9. **Never print\u002Fecho the Gemini key** in logs or commands.\n\n## Code & tests\n\n- **Executor (SDK driver):** `omnigent\u002Finner\u002Fantigravity_executor.py`\n- **Wrap (HARNESS_ANTIGRAVITY_* env → executor):** `omnigent\u002Finner\u002Fantigravity_harness.py`\n- **Auth \u002F key resolution:** `omnigent\u002Fonboarding\u002Fantigravity_auth.py`\n- **Spawn env:** `_build_antigravity_spawn_env` in `omnigent\u002Fruntime\u002Fworkflow.py`\n\n```bash\n# Unit tests (use --frozen; the cwsandbox extra is unsatisfiable on public PyPI here)\nuv run --frozen --group test python -m pytest \\\n  tests\u002Finner\u002Ftest_antigravity_executor.py \\\n  tests\u002Finner\u002Ftest_antigravity_harness.py \\\n  tests\u002Fruntime\u002Ftest_antigravity_spawn_env.py \\\n  tests\u002Fonboarding\u002Ftest_antigravity_auth.py -q\n# (or, if uv re-resolve is blocked on your host: .venv\u002Fbin\u002Fpython -m pytest \u003Csame paths> -q)\n```\n\nThere is no gated per-harness antigravity e2e test yet (it is deliberately\nexcluded from the live no-AGENT harness matrix in\n`tests\u002Fe2e\u002Fomnigent\u002Ftest_run_harness_without_agent_e2e.py`, because that matrix\nauthenticates through the Databricks gateway and antigravity is Gemini-native).\nThis skill IS the live coverage.\n\n## Bug-bash (fan out)\n\nTo stress the harness, run several scenario probes in parallel — each builds a\nbundle and runs real turns against the same `$SERVER`, then reports what broke.\nHighest-value targets: the `custom_tools` bridge (hangs \u002F lost tool results \u002F\nerrors reported as success), model routing, policy enforcement, streamed-output\nrendering, history retention across turns, and orphaned `localharness`\nprocesses after teardown.\n\n## Known sharp edges (found via the merge review — \"as of this writing\")\n\nSeveral were merged as-is and have **fix PRs in flight (#276–#281)** — verify\nagainst your checkout:\n\n- **Native\u002Fbuilt-in tools bypass the TOOL_CALL policy.** Only a\n  `PostToolCallHook` (post-execution, can't block) was installed at merge, so a\n  DENY\u002FASK guardrail doesn't gate the SDK's native shell\u002Ffile tools before they\n  run. Bridged `sys_*` tools route through the server. *(Fix: policy-enforcement PR.)*\n- **LLM_REQUEST \u002F LLM_RESPONSE policies aren't evaluated** in `run_turn` (prompt-\n  deny \u002F output-block silently ignored). *(Fix: policy-enforcement PR.)*\n- **History on a fresh\u002Frebuilt session.** The SDK has no history-injection API,\n  so prior turns are replayed as a plain-text `\"Conversation so far: …\"` prefix\n  (user\u002Fassistant text only; tool calls aren't reconstructed). *(PR #278.)*\n- **`sys_list_models` can over-report OpenAI-family models** for antigravity\n  (it was mapped to the openai family for shared lookups); the worker only runs\n  Gemini. *(Fix: openai-family-cleanup PR.)*\n- **Per-session `\u002Fmodel` override** was rejected with a false \"no plumbing\"\n  error. *(PR #276.)*  **Global `auth:` (an OpenAI key)** could be adopted as a\n  Gemini key. *(PR #277.)*  **Tool parameter schemas** were dropped (model flew\n  blind on arg shapes). *(PR #279.)*\n- **A failed turn** (e.g. the glibc error, a bad model) surfaces as a `failed`\n  session + an error item — if a turn returns little, check\n  `GET \u002Fv1\u002Fsessions\u002F{id}` status and `…\u002Fitems` rather than assuming success.\n\n## Cleanup\n\n```bash\n.venv\u002Fbin\u002Fomni server stop      # stop the managed background server\nrm -rf \u002Ftmp\u002Fagy-dev             # remove scratch bundles\npgrep -af \"localharness\"        # confirm no orphaned native subprocesses linger\n```\n","# Antigravity SDK harness: end-to-end dev & testing\n\nThe `antigravity` harness drives Google's **Antigravity Python SDK**\n(`google-antigravity`, an in-process `Agent`\u002F`Conversation`) and bridges\nOmnigent's `sys_*` tools into the SDK as `custom_tools`. It is **Gemini-native**:\nit authenticates with a Gemini \u002F Antigravity API key (or Vertex AI) and has **no\nOpenAI-compatible gateway \u002F Databricks path**. This skill is the proven recipe\nfor running it **for real** against a live local server — not just the unit\ntests.\n\n> The harness runs as a **local runner** from your current checkout, so\n> `omni run \u003Cbundle> --server \u003Curl>` exercises exactly the code you're on.\n\n## Prerequisites (check these first)\n\n1. **You're on the branch you want to test.** The antigravity harness merged to\n   `main` (#194). Test on `main` unless validating a specific branch.\n2. **A Gemini API key is configured.** The SDK *requires* one (`AIza…`); there\n   is no login flow. Verify (booleans only — never print the key):\n   ```bash\n   .venv\u002Fbin\u002Fpython -c \"from omnigent.onboarding.antigravity_auth import antigravity_api_key_configured as c; import os; print('config:', c(), 'env:', bool(os.environ.get('GEMINI_API_KEY') or os.environ.get('ANTIGRAVITY_API_KEY')))\"\n   ```\n   If both are `False`, run `omni setup` → **Antigravity** and paste a key, or\n   `export GEMINI_API_KEY=AIza…`.\n3. **`google-antigravity` is installed** (the `antigravity` extra —\n   `pip install \"omnigent[antigravity]\"`):\n   `.venv\u002Fbin\u002Fpython -c \"import google.antigravity as a; print(a.__file__)\"`.\n4. **glibc ≥ ~2.36.** The SDK spawns a **native `localharness` binary** that\n   needs a recent glibc (`GLIBC_ABI_DT_RELR`). Check `ldd --version | head -1`.\n   On an older host the turn fails at setup with\n   `RuntimeError: … localharness: … version 'GLIBC_ABI_DT_RELR' not found`. Dev\n   workaround on a glibc-2.31 box: point the SDK at a loader-shim via\n   `ANTIGRAVITY_HARNESS_PATH=\u002Fpath\u002Fto\u002Fshim` that runs the *untouched* bundled\n   binary through a newer glibc's loader (see the auto-memory note\n   `antigravity-harness-glibc-native-binary.md`). The shim is dev-only — the\n   real fix is a glibc-≥2.36 host.\n5. **Network egress to the Gemini backend.** The native binary talks to\n   Google's API; a turn that hangs or fails to connect on a locked-down host is\n   usually an egress problem, not a harness bug.\n\n## Step 1 — start a local server\n\n```bash\ncd \u002Fpath\u002Fto\u002Fomnigent\n.venv\u002Fbin\u002Fomni server --background          # spawns a detached server on a free loopback port\n.venv\u002Fbin\u002Fomni server status         # prints the URL, e.g. http:\u002F\u002F127.0.0.1:6767\n```\n\nUse the **printed URL** below as `$SERVER`. (You can also run a foreground\nserver on a fixed port with `omnigent server --port 7777 --no-open`.)\n\n## Step 2 — build an antigravity agent bundle\n\nA spec with `spec_version` **must be a directory containing `config.yaml`** —\nnot a single `.yaml` file. Minimal antigravity agent (no `auth:` block → it\nresolves the key from the `antigravity:` config \u002F ambient env):\n\n```bash\nmkdir -p \u002Ftmp\u002Fagy-dev\ncat > \u002Ftmp\u002Fagy-dev\u002Fconfig.yaml \u003C\u003C'YAML'\nspec_version: 1\nname: agy-dev\ndescription: Antigravity SDK dev\u002Ftest agent.\nexecutor:\n  type: omnigent\n  config:\n    harness: antigravity\n    model: gemini-3.5-flash      # default; gemini-3-pro 404s on a plain AI-Studio key\nprompt: |\n  You are a terse test agent. Answer in as few words as possible.\nYAML\n```\n\nFor sub-agents, tools, guardrails\u002Fpolicies, copy the field shapes from\n`examples\u002Fpolly\u002Fconfig.yaml` and `examples\u002Fdebby\u002Fconfig.yaml`.\n\n## Step 3 — run a turn (and smoke-test)\n\n```bash\nSERVER=http:\u002F\u002F127.0.0.1:6767   # the URL from `omni server status`\ntimeout 280 .venv\u002Fbin\u002Fomni run \u002Ftmp\u002Fagy-dev \\\n  -p \"Reply with exactly the single word: PONG\" \\\n  --server \"$SERVER\" 2>&1\n```\n\nA healthy run prints connection lines then the assistant reply (`PONG`). If\nthat works, the full stack is good: Gemini key, glibc\u002Fnative binary, egress,\nstreaming, harness.\n\n- **Shell \u002F file tools:** add `--tools coding`.\n- **Specific model:** add `--model gemini-2.5-flash` (or another Gemini id).\n\n## Targeted scenarios\n\n| Goal | How |\n|------|-----|\n| Native tools (shell\u002Fedit\u002Fread) | `--tools coding`, prompt to create→read→edit a file and run a shell command; confirm it actually touches disk |\n| Bridged `sys_*` \u002F sub-agent dispatch | declare a sub-agent (`tools.agents`\u002F`spawn`), prompt the agent to delegate — exercises the `custom_tools` bridge + `PostToolCallHook` |\n| Model routing | run the same bundle with several `--model` Gemini ids; note which actually runs |\n| Vertex AI auth | set `executor.config.vertex: true` + `project`\u002F`location` and use GCP application-default creds instead of an API key |\n| Policy \u002F guardrail | add a guardrail that denies a keyword; confirm it blocks (see the **sharp edges** below — LLM-phase + tool-call enforcement was incomplete at merge) |\n| Per-session brain override | run a bundle agent (polly\u002Fdebby) and select `antigravity` as the brain harness (it's in `BRAIN_HARNESS_LABELS`) |\n| Concurrency \u002F leaks | fire several `omni run … &` at once; then `pgrep -af localharness` to check for orphaned native subprocesses |\n\n## Gotchas (these cost real time)\n\n1. **`config.yaml`'s `server:` defaults to a *remote* server.** Omitting\n   `--server` sends your turn to that remote deploy — which may be **stale** and\n   reject the antigravity harness with `executor.config.harness: must be one of\n   […], got 'antigravity'`. **Always pass `--server http:\u002F\u002F127.0.0.1:\u003Cport>`**\n   for local testing. (That allowlist is `omnigent\u002Fspec\u002F_omnigent_compat.py`; if\n   a *local* server rejects `antigravity`, it's running stale code — restart it\n   from your checkout.)\n2. **A spec with `spec_version` must be a directory + `config.yaml`**, never a\n   single `.yaml` file.\n3. **Antigravity needs a Gemini key** (no login). Resolution precedence: spec\n   `executor.auth` (api_key) > stored `antigravity:` config block (`omni setup`)\n   > ambient `GEMINI_API_KEY` \u002F `ANTIGRAVITY_API_KEY`. Vertex AI is opt-in via\n   `executor.config` `vertex`\u002F`project`\u002F`location`.\n4. **No OpenAI gateway \u002F Databricks.** The SDK has no `base_url`; a `databricks`\n   or generic-`provider` auth is **warned and ignored**, and the run falls back\n   to ambient Gemini creds. Don't expect `databricks-*` models to route through\n   the AI Gateway like claude-sdk\u002Fcodex\u002Fpi.\n5. **Model ids are Gemini ids.** Default `gemini-3.5-flash`. `gemini-3-pro`\n   **404s on a plain AI-Studio key** — use `gemini-2.5-flash` \u002F `gemini-3.5-flash`\n   unless your key has Pro access.\n6. **The native binary needs glibc ≥ ~2.36** (see Prereq 4). This is the most\n   common \"it won't even start\" cause; check it before assuming a harness bug.\n7. **Turns take ~10–60s** — always wrap in `timeout 280`.\n8. **Local-runner topology:** `omni run \u003Cbundle> --server \u003Curl>` runs the\n   harness from your **current checkout**; the server only holds state. The\n   managed `omni server --background` server runs from whatever venv launched it.\n9. **Never print\u002Fecho the Gemini key** in logs or commands.\n\n## Code & tests\n\n- **Executor (SDK driver):** `omnigent\u002Finner\u002Fantigravity_executor.py`\n- **Wrap (HARNESS_ANTIGRAVITY_* env → executor):** `omnigent\u002Finner\u002Fantigravity_harness.py`\n- **Auth \u002F key resolution:** `omnigent\u002Fonboarding\u002Fantigravity_auth.py`\n- **Spawn env:** `_build_antigravity_spawn_env` in `omnigent\u002Fruntime\u002Fworkflow.py`\n\n```bash\n# Unit tests (use --frozen; the cwsandbox extra is unsatisfiable on public PyPI here)\nuv run --frozen --group test python -m pytest \\\n  tests\u002Finner\u002Ftest_antigravity_executor.py \\\n  tests\u002Finner\u002Ftest_antigravity_harness.py \\\n  tests\u002Fruntime\u002Ftest_antigravity_spawn_env.py \\\n  tests\u002Fonboarding\u002Ftest_antigravity_auth.py -q\n# (or, if uv re-resolve is blocked on your host: .venv\u002Fbin\u002Fpython -m pytest \u003Csame paths> -q)\n```\n\nThere is no gated per-harness antigravity e2e test yet (it is deliberately\nexcluded from the live no-AGENT harness matrix in\n`tests\u002Fe2e\u002Fomnigent\u002Ftest_run_harness_without_agent_e2e.py`, because that matrix\nauthenticates through the Databricks gateway and antigravity is Gemini-native).\nThis skill IS the live coverage.\n\n## Bug-bash (fan out)\n\nTo stress the harness, run several scenario probes in parallel — each builds a\nbundle and runs real turns against the same `$SERVER`, then reports what broke.\nHighest-value targets: the `custom_tools` bridge (hangs \u002F lost tool results \u002F\nerrors reported as success), model routing, policy enforcement, streamed-output\nrendering, history retention across turns, and orphaned `localharness`\nprocesses after teardown.\n\n## Known sharp edges (found via the merge review — \"as of this writing\")\n\nSeveral were merged as-is and have **fix PRs in flight (#276–#281)** — verify\nagainst your checkout:\n\n- **Native\u002Fbuilt-in tools bypass the TOOL_CALL policy.** Only a\n  `PostToolCallHook` (post-execution, can't block) was installed at merge, so a\n  DENY\u002FASK guardrail doesn't gate the SDK's native shell\u002Ffile tools before they\n  run. Bridged `sys_*` tools route through the server. *(Fix: policy-enforcement PR.)*\n- **LLM_REQUEST \u002F LLM_RESPONSE policies aren't evaluated** in `run_turn` (prompt-\n  deny \u002F output-block silently ignored). *(Fix: policy-enforcement PR.)*\n- **History on a fresh\u002Frebuilt session.** The SDK has no history-injection API,\n  so prior turns are replayed as a plain-text `\"Conversation so far: …\"` prefix\n  (user\u002Fassistant text only; tool calls aren't reconstructed). *(PR #278.)*\n- **`sys_list_models` can over-report OpenAI-family models** for antigravity\n  (it was mapped to the openai family for shared lookups); the worker only runs\n  Gemini. *(Fix: openai-family-cleanup PR.)*\n- **Per-session `\u002Fmodel` override** was rejected with a false \"no plumbing\"\n  error. *(PR #276.)*  **Global `auth:` (an OpenAI key)** could be adopted as a\n  Gemini key. *(PR #277.)*  **Tool parameter schemas** were dropped (model flew\n  blind on arg shapes). *(PR #279.)*\n- **A failed turn** (e.g. the glibc error, a bad model) surfaces as a `failed`\n  session + an error item — if a turn returns little, check\n  `GET \u002Fv1\u002Fsessions\u002F{id}` status and `…\u002Fitems` rather than assuming success.\n\n## Cleanup\n\n```bash\n.venv\u002Fbin\u002Fomni server stop      # stop the managed background server\nrm -rf \u002Ftmp\u002Fagy-dev             # remove scratch bundles\npgrep -af \"localharness\"        # confirm no orphaned native subprocesses linger\n```\n",{"name":5,"description":7},"deploy\u002Froland-id-au-omniagent","83ee3dcd78468bbe3baf4bb264a0c99a53aa0fec",".claude\u002Fskills\u002Fantigravity-sdk-e2e-dev","https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent","https:\u002F\u002Fgithub.com\u002Froland-id-au","2026-09-16 03:15:34",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add roland-id-au\u002Fomniagent --skill antigravity-sdk-e2e-dev",{"tool":36,"command":37},"bunx","bunx skills add roland-id-au\u002Fomniagent --skill antigravity-sdk-e2e-dev",{"tool":39,"command":40},"pnpm","pnpm dlx skills add roland-id-au\u002Fomniagent --skill antigravity-sdk-e2e-dev","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent\u002Fblob\u002Fdeploy\u002Froland-id-au-omniagent\u002F.claude\u002Fskills\u002Fantigravity-sdk-e2e-dev\u002FSKILL.md",true,[],[45,52,59,66,71,76],{"skillNo":46,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":19},"01M7CTDVNPM78AJ051GVDB44X5","deep-research","Research checklist and source-quality guidance for the archer e2e fixture.","https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent\u002Fblob\u002Fdeploy\u002Froland-id-au-omniagent\u002Fwork\u002Fomnigent-main\u002Ftests\u002Fresources\u002Fexamples\u002Farcher\u002Fskills\u002Fdeep-research\u002FSKILL.md",[],[],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":19},"01M7CTDS8SPN8P5GK1Q4SCSY7K","pi-e2e-xyz-greet-c4a8d5","Test greeter — pi skills e2e. The unique suffix c4a8d5 appears in the skill name so a string match in the agent's output proves Pi loaded this skill.","https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent\u002Fblob\u002Fdeploy\u002Froland-id-au-omniagent\u002Fwork\u002Fomnigent-main\u002Ftests\u002Fresources\u002Fagents\u002Fpi_skills_none\u002Fskills\u002Fpi-e2e-xyz-greet-c4a8d5\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":19},"01M7CTDP4R9F06966W78WNZS2Y","pi-e2e-xyz-count-d2f6e1","Test counter — pi skills e2e. The unique suffix d2f6e1 appears in the skill name so a string match in the agent's output proves Pi loaded this skill.","https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent\u002Fblob\u002Fdeploy\u002Froland-id-au-omniagent\u002Fwork\u002Fomnigent-main\u002Ftests\u002Fresources\u002Fagents\u002Fpi_skills_none\u002Fskills\u002Fpi-e2e-xyz-count-d2f6e1\u002FSKILL.md",[],[],{"skillNo":67,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M7CTDKP4W1Q18ZCY0XANDSEG","https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent\u002Fblob\u002Fdeploy\u002Froland-id-au-omniagent\u002Fwork\u002Fomnigent-main\u002Ftests\u002Fresources\u002Fagents\u002Fpi_skills_list\u002Fskills\u002Fpi-e2e-xyz-greet-c4a8d5\u002FSKILL.md",[],[],{"skillNo":72,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M7CTDFBMA3EBZNV7J380THGP","https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent\u002Fblob\u002Fdeploy\u002Froland-id-au-omniagent\u002Fwork\u002Fomnigent-main\u002Ftests\u002Fresources\u002Fagents\u002Fpi_skills_list\u002Fskills\u002Fpi-e2e-xyz-count-d2f6e1\u002FSKILL.md",[],[],{"skillNo":77,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M7CTDCVTXTG72BKEV2HY7M6W","https:\u002F\u002Fgithub.com\u002Froland-id-au\u002Fomniagent\u002Fblob\u002Fdeploy\u002Froland-id-au-omniagent\u002Fwork\u002Fomnigent-main\u002Ftests\u002Fresources\u002Fagents\u002Fpi_skills_all\u002Fskills\u002Fpi-e2e-xyz-greet-c4a8d5\u002FSKILL.md",[],[]]