Pydantic AI Harness denial of service via ANSI stripping in ToolOutputLimits
Affects pydantic-ai-harness versions 0.4.0 through 0.69.0 in applications that run tools whose output an attacker can influence. The checked facts list no fixed version.
- Exploited
- No known exploitation
- Fix
- See vendor advisory
- Who is exposed
- Not reported
Summary
The pydantic-ai-harness package's ToolOutputLimits capability can be made to stall the event loop when ANSI stripping is enabled, delaying other requests and agent runs in the same process. An attacker who can place text in content a tool returns, such as shell output, file contents, fetched pages, or MCP server responses, can trigger this remotely without logging in or any user action.
Written by a model
Original description
Summary ui
With strip_ansi=True, Pydantic AI Harness's ToolOutputLimits capability removes ANSI escape sequences from tool output with a pattern that takes time quadratic in the length of a line. A tool result of a few hundred kilobytes, crafted with many unterminated escape sequences, can occupy the event loop for tens of seconds to minutes, delaying every other request and agent run in the process.
Details
This affects applications that set strip_ansi=True on ToolOutputLimits (named OverflowingToolOutput before 0.13.0) and run tools whose output an attacker can influence, such as shell commands, file reads, fetched pages or MCP servers. Stripping runs on the event loop before the capability reduces the output, so its own size limits do not bound it.
The assumed attacker can place text in content a tool returns.
Mitigation
Upgrade pydantic-ai-harness to 0.70.0 or later. Until then, leave strip_ansi at its default, False.
On Pydantic AI v1, the harness cannot be upgraded past 0.4.0. Upgrade to Pydantic AI v2 and pydantic-ai-harness 0.70.0, or leave strip_ansi off.
What to do
Workaround: The checked facts give no workaround. Limiting the application to tools whose output an attacker cannot influence avoids the setup described in the advisory.
Check your systems with an agent
Paste this into a coding or ops agent. It checks your systems read-only and reports a table per asset.
Check whether my systems are affected by CVE-2026-133161 (Harness `ToolOutputLimits` ANSI stripping can take quadratic time, blocking the event loop), using the cve.now data below.
Rules: work read-only. Do not run exploit code, proof-of-concept scripts or scanners that send attack payloads, and do not change any configuration, package, image or deployment. Report findings only.
What cve.now has for CVE-2026-133161 (entry last updated 2026-11-19):
Affected:
- PyPI package pydantic-ai-harness: from 0.4.0, through 0.69.0
Fixed versions: none listed yet.
Conditions (all must be true for the bug to be hit):
- Affected pydantic-ai-harness versions 0.4.0 through 0.69.0 (source: https://github.com/pydantic/pydantic-ai/security/advisories/GHSA-m4cr-m4qc-9hh3)
- Application runs tools whose output an attacker can influence, such as shell commands, file reads, fetched pages or MCP servers (source: https://github.com/pydantic/pydantic-ai/security/advisories/GHSA-m4cr-m4qc-9hh3)
Setup: unknown whether the bug needs a non-default feature or setting. Check the advisory.
Attacker needs: network access, no login, no user action.
CISA KEV: CVE-2026-133161 is not listed as of cve.now's last read.
This KEV status is for CVE-2026-133161 only. If you also check another CVE, look up its own KEV status; never assume it from this one.
How to check:
- Packages (PyPI): find every lockfile in each project and run the stack.check operation of the cve.now MCP server (execute) on each one. Compare the resolved versions, not the ranges a manifest declares, with the affected versions above. Include transitive dependencies.
Verify against the linked advisories before concluding, and cite the date you read each one:
- https://github.com/pydantic/pydantic-ai/security/advisories/GHSA-m4cr-m4qc-9hh3
Do not conclude an asset is unaffected unless you checked it and the evidence covers both its version and the conditions above. If you could not check something, mark it unknown and say what is missing.
Report a table with one row per asset and these columns: Asset | Version or build | Status (affected, fixed, not applicable, unknown) | Compensating controls | Exposure (internet-facing, internal, none) | Evidence | Next action.
Entry: https://cve.now/CVE-2026-133161
For follow-up questions, use the cve.now MCP server at https://mcp.cve.now/mcp (no authentication), for example execute with name cve.get and arguments {"id": "CVE-2026-133161"}.Affected versions
cve.now Index
- Exploitation0/40No signal
- Exposure20/25No login or user action needed, CISA rates it automatable · Unknown: Advisories don't say whether default setups are affected
- Reach16/20pydantic-ai-harness: 1.3M downloads a week
- Severity7.5/10CVSS 7.5
- No patch5/5No fixed version listed; see vendor advisory
What would change the score
- Would score about 94 if added to CISA KEV
- Would score about 89 if CISA's SSVC rates exploitation active
- Would score about 54 if advisories confirm default setups are exposed
Impact
- Widely usedpydantic-ai-harness has 1.3m weekly downloads. pypistats.org
Timeline
- CVE published
- First seen by cve.now
- CISA SSVC assessment
Exploitation: none · Automatable: yes · Technical impact: partial
- Record last updated
Evidence 3
| Kind | Source |
|---|---|
| Vendor advisory 1Establishes: Affected and fixed versions. | |
| Advisory | GitHub · pydantic/pydantic-ai · Advisory GHSA-m4cr-m4qc-9hh3 |
| CVE record 2Establishes: The identifier and description. | |
| Record | CVE List · CVE-2026-133161 |
| Record | NVD · CVE-2026-133161 |
Questions about CVE-2026-133161
Is CVE-2026-133161 being exploited?
It is not in CISA's Known Exploited Vulnerabilities catalog, and CISA's SSVC assessment does not mark exploitation as active. That does not mean it has not been attacked: vendors and researchers can report attacks before CISA lists them. No public exploit code is tracked.
Source: cve.now feed data
What does this mean?
ToolOutputLimits ANSI stripping can take quadratic time on crafted tool output of a few hundred kilobytes. This can occupy the event loop for tens of seconds to minutes, delaying other requests and agent runs in the same process.
Source: github.com
Which versions are affected by CVE-2026-133161?
pydantic-ai-harness (PyPI): from 0.4.0 through 0.69.0.
Source: cve.now feed data
Am I affected, and how do I check?
Affected versions are pydantic-ai-harness 0.4.0 through 0.69.0. The setup is an application that runs tools whose output an attacker can influence, such as shell commands, file reads, fetched pages or MCP servers. Check the installed version with pip show pydantic-ai-harness.
Source: github.com
How do I fix CVE-2026-133161?
The checked facts name no fixed version and no vendor patch identifiers, so no upgrade target can be given from them. Check the advisory for an update.
Source: github.com
What if there's no patch?
The checked facts give no workaround. Limiting the application to tools whose output an attacker cannot influence avoids the setup described in the advisory.
Source: github.com
Related
- CVE-2026-114386Pydantic AI Harness before 0.59.0 CodeMode Tool Call Limit Bypass via run_code
- CVE-2026-84381HTTPX2: Secure WebSocket traffic sent without TLS through SOCKS proxies
- CVE-2026-107295`pydantic-ai-slim` web UI `/api/chat` accepts browser-simple cross-origin requests that can trigger agent tool execution
- CVE-2026-107286Pydantic AI: Concurrency-limited models can keep their slot when a streamed request ends early
- CVE-2026-84382HTTPX2: Streaming response decompression does not bound peak memory (decompression amplification)
- CVE-2026-107289Pydantic AI: SSRF cloud-metadata blocklist bypass via IPv6 zone identifier (incomplete fix for CVE-2026-46678 and CVE-2026-48782)
No corrections so far. Something wrong on this page? Report an error.