Back to skill

Security audit

gpt-image-2.5

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent RunAPI image-generation recipe with a disclosed third-party CLI dependency and no evidence of hidden or destructive behavior.

Before installing, be comfortable trusting the RunAPI Homebrew tap and CLI with the local media files you choose to process and any RunAPI credentials you provide. Prefer a narrowly scoped API key, avoid exposing unrelated secrets in the environment, and consider pinning or otherwise verifying the CLI release in stricter environments.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:15
Finding
Unpinned Third-Party CLI Dependency## Vulnerability Details **File Location**: `SKILL.md`, lines 15–19 **Vulnerability Type**: Supply-chain risk from an unpinned third-party executable **Risk Level**: Medium **Complete Code Snippet**: ```yaml install: - kind: brew formula: runapi-ai/tap/runapi bins: - runapi ``` ### Technical Analysis The skill instructs the environment to install the `runapi` executable from the third-party Homebrew tap `runapi-ai/tap` without pinning a reviewed version, immutable formula revision, checksum, or publisher signature. The installed executable is subsequently trusted to authenticate with RunAPI, process local media paths, communicate over the network, and submit potentially billable API tasks. Because the dependency is mutable and its implementation is not included in the audited project, later upstream changes could alter the executable behavior after this skill has been reviewed. This is a supply-chain weakness rather than evidence that the current upstream package is malicious. ### Attack Path 1. An attacker compromises the third-party Homebrew tap, its release infrastructure, or an authorized publisher account. 2. The attacker modifies the mutable formula or referenced release so that it installs a malicious `runapi` executable. 3. A user or agent follows the skill installation configuration and resolves the compromised dependency. 4. The malicious executable runs during authentication, contract discovery, file processing, or task submission. 5. It can access data available to the invoking process, potentially including RunAPI credentials, saved CLI authentication state, selected local media, and API request contents. 6. It may exfiltrate those assets, tamper with requests or results, or initiate unauthorized API operations within the permissions of the available credentials. ### Impact Assessment Successful exploitation would execute code with the privileges of the user running the skill. The poten ...[truncated 651 chars]
Remediation
## Remediation Suggestions 1. Pin the CLI to a specific reviewed release and, where supported, an immutable Homebrew formula or tap commit. 2. Publish and verify cryptographic checksums or publisher signatures for the CLI artifact before installation or execution. 3. Document the canonical source repository and trusted release channel so package provenance can be verified. 4. Add a controlled update process that reviews dependency changes before advancing the pinned version. 5. Run the CLI with least privilege in a restricted environment that exposes only the required input and output files. 6. Provide narrowly scoped credentials through process-local environment injection, and avoid exposing unrelated secrets. 7. Restrict outbound network access to the documented RunAPI endpoints where the runtime supports network allowlisting. 8. Prefer an installation mechanism that supports lockfiles, reproducible artifacts, and integrity verification.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Static analysis

No suspicious patterns detected.