PYSEC-2026-3875
Vulnerability from pysec - Published: 2026-09-10 09:44 - Updated: 2026-09-10 11:02Summary
An authenticated, non-admin user can obtain arbitrary host-filesystem read/write (and host environment-secret disclosure) on an Omnigent runner by uploading an agent bundle whose os_env.cwd points outside any intended workspace (e.g. / or /home/<victim>). The cwd field is taken verbatim from the bundle with no validation, normalization, or boundary check anywhere in the spec pipeline.
This is a different sink from GHSA-jrrm-9hc7-2v3h (CWE-94, shared-agent bundle overwrite -> stdio MCP RCE). It shares the bundle-upload vector but is reached through the user's own session-scoped agent and is not addressed by that advisory's proposed shared-agent guard.
Preconditions
- Runner realizes a session-scoped uploaded bundle without
OMNIGENT_RUNNER_WORKSPACEset. When that env var is set (CLI- and host-launched sessions set it), the speccwdis overridden and the attack is neutralized — so this is deployment-gated, not universal. - Attacker is any authenticated user (no admin scope;
_require_useronly checks identity). No shared-agent overwrite needed.
Details (verified against code)
- Parse — no validation.
omnigent/spec/parser.py:696storescwd=str(cwd_raw)verbatim. Absolute paths (/,/etc),../.., etc. are all accepted. Thesandbox.typeis likewise author-chosen and"none"is legal. - Validate — cwd unconstrained.
omnigent/spec/validator.py_validate_os_envchecks only fork/scratch/egress combinations; it never referencescwd(the sole mention, line ~526, is a comment). No boundary is applied to the cwd itself. The boundary inserver/schemas.pyvalidates a caller-supplied workspace against the spec cwd (treating the author cwd as trusted) and only for host-launched sessions — it does not bound the cwd. - Sink.
omnigent/inner/os_env.py:890setscwd = Path(spec.cwd or os.getcwd()).resolve(strict=False)as the environment root;os_env.py:934doesshutil.copytree(src=cwd, ...)whenfork=true. All agent file/shell tools are bounded by_assert_within_cwd(os_env.py:1040), which checksresolved.relative_to(cwd)— but since cwd is attacker-controlled,cwd=/makes the entire host filesystem in-bounds for read and write;fork=truewithcwd=/home/victimcopies that tree into the agent-readable workspace. - Decisive gate.
omnigent/runner/resource_registry.py:648-654:cwd = default_cwdonly whenself._runner_workspace is not None or spec_os_env.cwd in (None, ".", "./"); otherwisecwd = spec_os_env.cwd(the attacker's absolute path). SoOMNIGENT_RUNNER_WORKSPACEis the only thing standing between the spec and the host FS — and it is an operational control, not an in-code guard. A code comment attool_dispatch.py:~4207claims cwd "is treated as a boundary at session-create time," which is not true on this path.
Attack path
- Authenticated user sends
POST /v1/sessions(multipart) with an agent bundle whoseconfig.yamlcontains:yaml os_env: cwd: "/" # or /home/<victim>, with fork: true for one-shot exfil sandbox: { type: none } - On a runner without
OMNIGENT_RUNNER_WORKSPACE, the agent'ssys_os_read/write/edit/shelltools now operate over the whole host filesystem, and (sandbox inactive) inherit the runner's full environment — exposing host secrets via e.g.sys_os_shell("env").
Impact
Arbitrary host-filesystem read/write and runner-environment secret disclosure, available to any authenticated user on an affected deployment. High severity (deployment-conditional).
Suggested fix
Add a control on the cwd field itself in omnigent/spec/_validate_os_env (and/or at parse): reject absolute paths and .. traversal, and require cwd to resolve within the runner workspace / an allow-listed root. Do not rely on OMNIGENT_RUNNER_WORKSPACE being set as the sole defense. Consider also disallowing bundle-author sandbox.type: none for server-realized (non-CLI) sessions.
Related
GHSA-jrrm-9hc7-2v3h (same bundle-upload vector, different sink; that fix does not cover this).
| Name | purl | omnigent | pkg:pypi/omnigent |
|---|
{
"affected": [
{
"package": {
"ecosystem": "PyPI",
"name": "omnigent",
"purl": "pkg:pypi/omnigent"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.3.0"
}
],
"type": "ECOSYSTEM"
}
],
"versions": [
"0.0.1rc1",
"0.0.1rc2",
"0.1.0",
"0.1.0rc1",
"0.1.0rc2",
"0.1.0rc3",
"0.1.0rc4",
"0.1.1",
"0.1.1rc2",
"0.2.0",
"0.2.0rc1",
"0.3.0rc1"
]
}
],
"aliases": [
"CVE-2026-62677",
"GHSA-p8rw-8qj3-hf33"
],
"details": "### Summary\n\nAn authenticated, non-admin user can obtain **arbitrary host-filesystem read/write** (and host environment-secret disclosure) on an Omnigent **runner** by uploading an agent bundle whose `os_env.cwd` points outside any intended workspace (e.g. `/` or `/home/\u003cvictim\u003e`). The `cwd` field is taken **verbatim** from the bundle with no validation, normalization, or boundary check anywhere in the spec pipeline.\n\nThis is a **different sink** from GHSA-jrrm-9hc7-2v3h (CWE-94, shared-agent bundle overwrite -\u003e stdio MCP RCE). It shares the bundle-upload *vector* but is reached through the user\u0027s **own** session-scoped agent and is **not** addressed by that advisory\u0027s proposed shared-agent guard.\n\n### Preconditions\n\n- Runner realizes a session-scoped uploaded bundle **without** `OMNIGENT_RUNNER_WORKSPACE` set. When that env var is set (CLI- and host-launched sessions set it), the spec `cwd` is overridden and the attack is neutralized \u2014 so this is deployment-gated, not universal.\n- Attacker is any authenticated user (no admin scope; `_require_user` only checks identity). No shared-agent overwrite needed.\n\n### Details (verified against code)\n\n1. **Parse \u2014 no validation.** `omnigent/spec/parser.py:696` stores `cwd=str(cwd_raw)` verbatim. Absolute paths (`/`, `/etc`), `../..`, etc. are all accepted. The `sandbox.type` is likewise author-chosen and `\"none\"` is legal.\n2. **Validate \u2014 cwd unconstrained.** `omnigent/spec/validator.py` `_validate_os_env` checks only fork/scratch/egress combinations; it never references `cwd` (the sole mention, line ~526, is a comment). No boundary is applied to the cwd itself. The boundary in `server/schemas.py` validates a *caller-supplied workspace against* the spec cwd (treating the author cwd as trusted) and only for host-launched sessions \u2014 it does not bound the cwd.\n3. **Sink.** `omnigent/inner/os_env.py:890` sets `cwd = Path(spec.cwd or os.getcwd()).resolve(strict=False)` as the environment root; `os_env.py:934` does `shutil.copytree(src=cwd, ...)` when `fork=true`. All agent file/shell tools are bounded by `_assert_within_cwd` (`os_env.py:1040`), which checks `resolved.relative_to(cwd)` \u2014 but since **cwd is attacker-controlled**, `cwd=/` makes the entire host filesystem in-bounds for read and write; `fork=true` with `cwd=/home/victim` copies that tree into the agent-readable workspace.\n4. **Decisive gate.** `omnigent/runner/resource_registry.py:648-654`: `cwd = default_cwd` only when `self._runner_workspace is not None or spec_os_env.cwd in (None, \".\", \"./\")`; otherwise `cwd = spec_os_env.cwd` (the attacker\u0027s absolute path). So `OMNIGENT_RUNNER_WORKSPACE` is the only thing standing between the spec and the host FS \u2014 and it is an operational control, not an in-code guard. A code comment at `tool_dispatch.py:~4207` claims cwd \"is treated as a boundary at session-create time,\" which is not true on this path.\n\n### Attack path\n\n1. Authenticated user sends `POST /v1/sessions` (multipart) with an agent bundle whose `config.yaml` contains:\n ```yaml\n os_env:\n cwd: \"/\" # or /home/\u003cvictim\u003e, with fork: true for one-shot exfil\n sandbox: { type: none }\n ```\n2. On a runner without `OMNIGENT_RUNNER_WORKSPACE`, the agent\u0027s `sys_os_read`/`write`/`edit`/`shell` tools now operate over the whole host filesystem, and (sandbox inactive) inherit the runner\u0027s full environment \u2014 exposing host secrets via e.g. `sys_os_shell(\"env\")`.\n\n### Impact\n\nArbitrary host-filesystem read/write and runner-environment secret disclosure, available to any authenticated user on an affected deployment. High severity (deployment-conditional).\n\n### Suggested fix\n\nAdd a control on the `cwd` field itself in `omnigent/spec/_validate_os_env` (and/or at parse): reject absolute paths and `..` traversal, and require `cwd` to resolve within the runner workspace / an allow-listed root. Do not rely on `OMNIGENT_RUNNER_WORKSPACE` being set as the sole defense. Consider also disallowing bundle-author `sandbox.type: none` for server-realized (non-CLI) sessions.\n\n### Related\n\nGHSA-jrrm-9hc7-2v3h (same bundle-upload vector, different sink; that fix does not cover this).",
"id": "PYSEC-2026-3875",
"modified": "2026-09-10T11:02:18.308444Z",
"published": "2026-09-10T09:44:59.689355Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/omnigent-ai/omnigent/security/advisories/GHSA-p8rw-8qj3-hf33"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-62677"
},
{
"type": "WEB",
"url": "https://github.com/omnigent-ai/omnigent/pull/1417"
},
{
"type": "WEB",
"url": "https://github.com/omnigent-ai/omnigent/commit/7ca0cca3c9a65c04c489edf68f0e080424a26868"
},
{
"type": "PACKAGE",
"url": "https://github.com/omnigent-ai/omnigent"
},
{
"type": "WEB",
"url": "https://github.com/omnigent-ai/omnigent/releases/tag/v0.3.0"
},
{
"type": "PACKAGE",
"url": "https://pypi.org/project/omnigent"
},
{
"type": "ADVISORY",
"url": "https://github.com/advisories/GHSA-p8rw-8qj3-hf33"
}
],
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H",
"type": "CVSS_V3"
}
],
"summary": "Omnigent: Unvalidated os_env.cwd in agent bundle yields arbitrary host filesystem access on runners without OMNIGENT_RUNNER_WORKSPACE"
}
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.