The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The skill frequently uses npx ruv-swarm to execute release tasks. This package is not from a trusted source/organization listed in the security profile. Running unverified packages via npx can lead to supply chain attacks where malicious code is downloaded and executed at runtime.
REMOTE_CODE_EXECUTION (MEDIUM): The reliance on npx for core functionality allows for the execution of remote code that has not been audited or pinned to a specific hash or version in the examples, providing a vector for arbitrary code execution.
COMMAND_EXECUTION (LOW): The skill makes extensive use of the GitHub CLI (gh) and shell scripting to perform sensitive operations like creating releases, uploading artifacts, and modifying repository state. While expected for a release tool, it represents a high-privilege capability surface.
Indirect Prompt Injection (LOW): The skill ingests untrusted data from GitHub (commit messages, PR titles) and passes them to the AI-driven ruv-swarm tool, which could be exploited to influence the agent's behavior during release generation.
Ingestion points: Commit messages and PR data are fetched via gh api and gh pr list (SKILL.md).
Boundary markers: Absent; untrusted strings are passed directly as command-line arguments to npx ruv-swarm.
Capability inventory: The skill can create GitHub releases, upload assets, create issues, and publish to npm.
Sanitization: No sanitization or validation of the commit/PR data is performed before processing.

github-release-swarm