Back to skill

Security audit

AKY-ES

Security checks for vulnerabilities and agentic risk

Overview

This skill is for Elasticsearch memory/search, but its local-only framing conflicts with automatic cloud fallback and unaudited external workspace scripts.

Review before installing. Use it only if you are comfortable with persistent searchable storage, a keepalive cron job, and possible Elastic Cloud use. Do not store sensitive material unless the external scripts are reviewed, the dependency is pinned, and cloud fallback is disabled or made explicit with destination confirmation.

Vulnerability Patterns
  • Tool Hijacking and SpoofingModifies or replaces tools so legitimate-looking calls execute attacker logic
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Findings (3)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:11
Finding

Unbounded Third-Party Dependency Version

Content
View full analysis
=9.4", }, ], ``` ### Technical Analysis The Skill permits installation of any Elasticsearch Python package version equal to or newer than 9.4. This open-ended constraint does not ensure that the installed artifact is the same version that was reviewed or tested. A future compromised, malicious, or incompatible upstream release could be selected automatically during installation. No package hash, lock file, trusted artifact digest, or upper version boundary is specified. Package installation also occurs before the Skill's documented functionality can be used, placing the package supply chain within the trusted execution path. This finding does not establish that the current `elasticsearch` package is malicious. The vulnerability is the absence of reproducible dependency controls. ### Attack Path 1. An attacker compromises the upstream package publication process, a maintainer account, or another component of the package distribution channel. 2. The attacker publishes a new version satisfying `elasticsearch>=9.4`. 3. A user installs or reinstalls the Skill in an environment where the malicious release is selected. 4. The package is installed into the Agent environment. 5. Malicious package behavior executes during installation or when the Skill imports and uses the dependency. ### Impact Assessment Successful exploitation could execute code with the permissions of the account performing installation or running the Skill. This may expose files, environment variables, Elasticsearch credentials, indexed documents, and other resources available to that account. The issue does not independently grant root privileges. Its scope is limited to the permissions and network access of the installation or r ...[truncated 23 chars]
Remediation
View remediation

T07 · Tool Hijacking and Spoofing

Error
Location
SKILL.md:10
Finding

Execution of Mutable External Workspace Tools Outside the Audited Package

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:56
Finding

Silent Fallback from Local Storage to Elastic Cloud

Content
View full analysis
cloud | | `ES_API_KEY` | API key (cloud) | — | | `ES_USER` | Basic auth user (cloud) | — | | `ES_PASS` | Basic auth password (cloud) | — | | `AUTH_MODE` | `api_key` or `basic` | `api_key` | ``` ### Technical Analysis The Skill is presented as a local Elasticsearch storage facility, but it states that an unreachable local endpoint causes automatic fallback to Elastic Cloud. Operations include storing conversations, reports, legal material, contracts, notes, and other long-term-memory content. This behavior crosses a material trust boundary: data intended for `localhost` may instead be transmitted to an external service. The documentation does not require per-operation consent, an explicit cloud-mode selection, destination confirmation, or data classification before fallback. The implementation is located in the absent external scripts, so TLS validation, endpoint allowlisting, credential handling, and the exact fallback conditions cannot be verified. The finding is limited to the unsafe behavior explicitly described by the Skill; it does not claim that data was actually transmitted during the audit. ### Attack Path 1. Cloud endpoint and authentication variables are present in the Agent environment. 2. Local Elasticsearch at `localhost:9200` is stopped, unreachable, misconfigured, or intentionally disrupted. 3. A user asks the Agent to save or bulk-import sensitive material under the expectation that storage is local. 4. The external tool detects the local connection failure and automatically selects Elastic Cloud. 5. The tool sends documents or queries to th ...[truncated 932 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill is presented as local-only, but the documentation states it will fall back to Elastic Cloud and supports cloud credentials. That creates a trust-boundary mismatch: data a user expects to stay on localhost may be transmitted to a remote service if local ES is unavailable, potentially exposing sensitive indexed content or queries.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The documented cloud fallback means requests and stored content may leave the local machine, but the skill does not prominently warn users about network transmission or credential use. In a persistence/search skill likely to handle notes, reports, and other sensitive material, silent fallback materially increases confidentiality risk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The environment variable section documents ES_URL, API key, username, password, and auth mode for cloud access, which contradicts the skill’s local-only description. This can mislead users and downstream agents into supplying secrets or using a remote endpoint under the assumption that the skill is confined to localhost.

Content

No source excerpt is available for this finding.