The Agent Skills Directory

[COMMAND_EXECUTION]: The provided webhook-handler.js integration example uses child_process.execSync with unsanitized input from GitHub issue comments (event.comment.body). An attacker could post a specially crafted comment (e.g., using shell metacharacters like ;, &&, or |) to execute arbitrary commands on the server hosting the webhook.
[REMOTE_CODE_EXECUTION]: The skill relies on npx to execute ruv-swarm for various orchestration tasks. While ruv-swarm is a resource associated with the author (ruvnet), the use of npx involves the dynamic execution of remote packages.
[EXTERNAL_DOWNLOADS]: The skill requires several external dependencies including the GitHub CLI (gh), ruv-swarm, and claude-flow to function.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from Pull Request diffs and comments to drive AI agent behavior.
Ingestion points: Data enters through gh pr view, gh pr diff, and webhook event bodies.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the command templates.
Capability inventory: The skill can perform sensitive actions including posting PR comments, requesting changes, adding labels, and merging code.
Sanitization: No evidence of sanitization or validation of the ingested PR content is present in the provided scripts.

github-code-review