Back to skill

Security audit

Migrate Model Defaults

Security checks for vulnerabilities and agentic risk

Overview

This skill is a narrowly scoped repo-editing helper with one disclosed supply-chain caution around an optional validation command.

Before installing, be aware that workflow validation may run actionlint through Go using an unpinned '@latest' dependency if actionlint is not already installed. Prefer a preinstalled or pinned actionlint version in sensitive repositories. The skill otherwise asks for confirmation before edits and stays focused on model-default migration.

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:50
Finding

Unpinned Remote Dependency Is Downloaded and Executed During Validation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 50
Vulnerability Type: Unpinned third-party tool execution
Risk Level: Medium

Vulnerable Code Snippet:

markdown
- For workflow edits, run `actionlint` if available or via `go run github.com/rhysd/actionlint/cmd/actionlint@latest <files>`.

Technical Analysis

The validation instructions use go run with the mutable @latest version selector. This causes Go tooling to retrieve source code from an external dependency source and execute whichever release is considered latest at invocation time.

Because the executed code is not pinned to an audited version or immutable commit, its effective contents can change after this skill has been reviewed. If the upstream repository, module distribution path, maintainer account, or a related supply-chain component is compromised, malicious code could be delivered through a future version and executed automatically during workflow validation.

This is classified as T08: Insecure Dependencies because the principal weakness is execution of an unpinned third-party dependency from an unsafe mutable source.

Attack Path

  1. An attacker compromises the upstream actionlint repository, a maintainer account, or another relevant part of the Go module distribution chain.
  2. The attacker publishes or causes a malicious release to be selected by the @latest version query.
  3. A user invokes this skill for edits involving workflow files.
  4. The agent follows the validation instruction and executes:
    shell
    go run github.com/rhysd/actionlint/cmd/actionlint@latest <files>
    
  5. Go downloads, compiles, and executes the attacker-controlled module code.
  6. The malicious code runs with the permissions and environmental access of the agent process.

Impact Assessment

Successful exploitation could execute arbitrary code under the account running the agent. Depending on the execution environment and sandbox restrictions, this ...[truncated 620 chars]

Remediation
View remediation

Remediation Suggestions

  1. Pin actionlint to a reviewed, immutable version rather than using @latest, for example:
    shell
    go run github.com/rhysd/actionlint/cmd/actionlint@vX.Y.Z <files>
    
  2. Review the selected release and commit before adopting it. Update the pinned version only through an explicit dependency-review process.
  3. Prefer a trusted, preinstalled actionlint binary when available, and verify its expected version before execution.
  4. Use Go module checksum verification and retain the expected dependency checksums in controlled project metadata where applicable.
  5. In sensitive environments, prohibit automatic network dependency retrieval during validation. Install the approved tool in a controlled build image or retrieve it through a trusted internal artifact registry.
  6. Run validation tools with least privilege, restricted filesystem access, no unnecessary credentials, and limited network connectivity.
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