GHSA-VCV2-R9JH-99M5
Vulnerability from github – Published: 2026-06-19 15:12 – Updated: 2026-06-19 15:12Summary
agentic-flow versions <= 2.0.13 MCP server tools interpolated attacker-influenceable tool parameters (e.g. agent, task, name, language, agentdb arguments) directly into shell command strings passed to execSync(). A malicious value reaching any of the affected MCP tools could break out of the surrounding double-quoted argument and execute arbitrary OS commands with the privileges of the user running the MCP server.
This was a partial-fix gap: prior commit 6a06854 (#158) fixed CWE-78 elsewhere in the project but missed the MCP server files entirely.
Impact
Any MCP tool argument that the AI agent treats as data but the implementation interpolates into a shell command string becomes a command-injection vector. In MCP deployments where untrusted content (web pages, files, third-party tool output) is processed by the agent, this is reachable without direct attacker access to the host. The HTTP/SSE transports (http-sse.ts, http-streaming-updated.ts) expose the same sinks without authentication or Origin/Host validation, which may raise the effective severity in any deployment that binds them to a reachable network interface.
Affected components
src/mcp/standalone-stdio.ts—agentic_flow_agent,agentic_flow_create_agent,agentic_flow_list_all_agents,agentic_flow_agent_info,agentic_flow_check_conflicts,agentic_flow_optimize_model,agentic_flow_list_agents,agent_booster_edit_file,agent_booster_batch_edit,agent_booster_parse_markdown,agentdb_stats,agentdb_pattern_store,agentdb_pattern_search,agentdb_pattern_statssrc/mcp/fastmcp/servers/claude-flow-sdk.tssrc/mcp/fastmcp/servers/stdio-full.tssrc/mcp/fastmcp/servers/http-streaming-updated.tssrc/mcp/fastmcp/servers/http-sse.tssrc/mcp/fastmcp/servers/poc-stdio.tssrc/mcp/fastmcp/tools/agent/{execute,list,parallel}.tssrc/mcp/fastmcp/tools/swarm/orchestrate.tssrc/mcp/fastmcp/tools/hooks/pretrain.ts(depth path only)
Proof of Concept
// Pre-fix (standalone-stdio.ts, agentic_flow_agent)
let cmd = `npx --yes agentic-flow --agent "${agent}" --task "${task}"`;
const result = execSync(cmd, { encoding: 'utf-8', ... });
Invoking the MCP tool with:
{
"agent": "coder",
"task": "x\"; touch /tmp/INJECTED; id > /tmp/rce.txt; echo \""
}
produces, after interpolation:
npx --yes agentic-flow --agent "coder" --task "x"; touch /tmp/INJECTED; id > /tmp/rce.txt; echo ""
When execSync hands that to /bin/sh -c, the shell parses three commands: the truncated npx, then touch /tmp/INJECTED, then id > /tmp/rce.txt; echo "". The marker file /tmp/INJECTED is created and the user's id output is written to /tmp/rce.txt.
Patches
Fixed in agentic-flow@2.0.14 — every affected call site rewritten to use execFileSync(file, argv, { shell: false }) so attacker-controlled argv elements are passed straight to execve(2) without shell parsing.
Fix PR: ruvnet/agentic-flow#170 (merged at 0c2ec96)
A regression test (tests/security/cwe-78-mcp-execsync.test.ts) was added that statically scans every src/mcp/**/*.ts file and fails the build if any new execSync() call is reintroduced outside of a documented exemption, plus a behavioural smoke check that the canonical PoC payload remains inert when passed as an argv element to execFileSync.
Workarounds
Upgrade to agentic-flow >= 2.0.14. There is no in-product configuration that mitigates this without upgrading.
Downstream pin
The ruflo / claude-flow / @claude-flow/cli packages bumped from 3.12.3 → 3.12.4 to pull the patched agentic-flow:
ruflo@3.12.4claude-flow@3.12.4@claude-flow/cli@3.12.4
End users running any of npx ruflo@latest, npx claude-flow@latest, or npx @claude-flow/cli@latest are pinned to the fixed version.
Credit
Reported by hackchang via a well-scoped red-team report package (npm_agentic-flow_report_package_20260618_163017.zip) that included a sink inventory, a minimized PoC payload, and a clear explanation of why this was a partial-fix gap rather than intended behaviour. The sink inventory directly drove the single-grep pass that closed every reachable call site; the PoC payload became the behavioural smoke test that proves the canonical attack stays inert as an argv element.
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 2.0.13"
},
"package": {
"ecosystem": "npm",
"name": "agentic-flow"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "2.0.14"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [],
"database_specific": {
"cwe_ids": [
"CWE-78"
],
"github_reviewed": true,
"github_reviewed_at": "2026-06-19T15:12:58Z",
"nvd_published_at": null,
"severity": "HIGH"
},
"details": "## Summary\n\n`agentic-flow` versions `\u003c= 2.0.13` MCP server tools interpolated attacker-influenceable tool parameters (e.g. `agent`, `task`, `name`, `language`, `agentdb` arguments) directly into shell command strings passed to `execSync()`. A malicious value reaching any of the affected MCP tools could break out of the surrounding double-quoted argument and execute arbitrary OS commands with the privileges of the user running the MCP server.\n\nThis was a partial-fix gap: prior commit `6a06854` (#158) fixed CWE-78 elsewhere in the project but missed the MCP server files entirely.\n\n## Impact\n\nAny MCP tool argument that the AI agent treats as data but the implementation interpolates into a shell command string becomes a command-injection vector. In MCP deployments where untrusted content (web pages, files, third-party tool output) is processed by the agent, this is reachable without direct attacker access to the host. The HTTP/SSE transports (`http-sse.ts`, `http-streaming-updated.ts`) expose the same sinks without authentication or Origin/Host validation, which may raise the effective severity in any deployment that binds them to a reachable network interface.\n\n## Affected components\n\n- `src/mcp/standalone-stdio.ts` \u2014 `agentic_flow_agent`, `agentic_flow_create_agent`, `agentic_flow_list_all_agents`, `agentic_flow_agent_info`, `agentic_flow_check_conflicts`, `agentic_flow_optimize_model`, `agentic_flow_list_agents`, `agent_booster_edit_file`, `agent_booster_batch_edit`, `agent_booster_parse_markdown`, `agentdb_stats`, `agentdb_pattern_store`, `agentdb_pattern_search`, `agentdb_pattern_stats`\n- `src/mcp/fastmcp/servers/claude-flow-sdk.ts`\n- `src/mcp/fastmcp/servers/stdio-full.ts`\n- `src/mcp/fastmcp/servers/http-streaming-updated.ts`\n- `src/mcp/fastmcp/servers/http-sse.ts`\n- `src/mcp/fastmcp/servers/poc-stdio.ts`\n- `src/mcp/fastmcp/tools/agent/{execute,list,parallel}.ts`\n- `src/mcp/fastmcp/tools/swarm/orchestrate.ts`\n- `src/mcp/fastmcp/tools/hooks/pretrain.ts` (depth path only)\n\n## Proof of Concept\n\n```ts\n// Pre-fix (standalone-stdio.ts, agentic_flow_agent)\nlet cmd = `npx --yes agentic-flow --agent \"${agent}\" --task \"${task}\"`;\nconst result = execSync(cmd, { encoding: \u0027utf-8\u0027, ... });\n```\n\nInvoking the MCP tool with:\n\n```json\n{\n \"agent\": \"coder\",\n \"task\": \"x\\\"; touch /tmp/INJECTED; id \u003e /tmp/rce.txt; echo \\\"\"\n}\n```\n\nproduces, after interpolation:\n\n```\nnpx --yes agentic-flow --agent \"coder\" --task \"x\"; touch /tmp/INJECTED; id \u003e /tmp/rce.txt; echo \"\"\n```\n\nWhen `execSync` hands that to `/bin/sh -c`, the shell parses three commands: the truncated `npx`, then `touch /tmp/INJECTED`, then `id \u003e /tmp/rce.txt; echo \"\"`. The marker file `/tmp/INJECTED` is created and the user\u0027s `id` output is written to `/tmp/rce.txt`.\n\n## Patches\n\nFixed in [`agentic-flow@2.0.14`](https://www.npmjs.com/package/agentic-flow/v/2.0.14) \u2014 every affected call site rewritten to use `execFileSync(file, argv, { shell: false })` so attacker-controlled argv elements are passed straight to `execve(2)` without shell parsing.\n\nFix PR: ruvnet/agentic-flow#170 (merged at `0c2ec96`)\n\nA regression test (`tests/security/cwe-78-mcp-execsync.test.ts`) was added that statically scans every `src/mcp/**/*.ts` file and fails the build if any new `execSync()` call is reintroduced outside of a documented exemption, plus a behavioural smoke check that the canonical PoC payload remains inert when passed as an argv element to `execFileSync`.\n\n## Workarounds\n\nUpgrade to `agentic-flow \u003e= 2.0.14`. There is no in-product configuration that mitigates this without upgrading.\n\n## Downstream pin\n\nThe `ruflo` / `claude-flow` / `@claude-flow/cli` packages bumped from `3.12.3` \u2192 `3.12.4` to pull the patched `agentic-flow`:\n\n- `ruflo@3.12.4`\n- `claude-flow@3.12.4`\n- `@claude-flow/cli@3.12.4`\n\nEnd users running any of `npx ruflo@latest`, `npx claude-flow@latest`, or `npx @claude-flow/cli@latest` are pinned to the fixed version.\n\n## Credit\n\nReported by **hackchang** via a well-scoped red-team report package (`npm_agentic-flow_report_package_20260618_163017.zip`) that included a sink inventory, a minimized PoC payload, and a clear explanation of why this was a partial-fix gap rather than intended behaviour. The sink inventory directly drove the single-grep pass that closed every reachable call site; the PoC payload became the behavioural smoke test that proves the canonical attack stays inert as an argv element.",
"id": "GHSA-vcv2-r9jh-99m5",
"modified": "2026-06-19T15:12:58Z",
"published": "2026-06-19T15:12:58Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/ruvnet/agentic-flow/security/advisories/GHSA-vcv2-r9jh-99m5"
},
{
"type": "WEB",
"url": "https://github.com/ruvnet/agentic-flow/issues/169"
},
{
"type": "WEB",
"url": "https://github.com/ruvnet/ruflo/issues/2414"
},
{
"type": "WEB",
"url": "https://github.com/ruvnet/agentic-flow/pull/170"
},
{
"type": "WEB",
"url": "https://github.com/ruvnet/ruflo/pull/2415"
},
{
"type": "PACKAGE",
"url": "https://github.com/ruvnet/agentic-flow"
},
{
"type": "WEB",
"url": "https://github.com/ruvnet/ruflo/releases/tag/v3.12.4"
},
{
"type": "WEB",
"url": "https://www.npmjs.com/package/agentic-flow/v/2.0.14"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
"type": "CVSS_V3"
}
],
"summary": "Agentic-Flow: OS Command Injection in agentic-flow MCP server tools via unsanitized tool-parameter interpolation into execSync"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
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.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
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.