The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes administrative tools like kubectl to perform deployment rollouts, scaling operations, and secret management. It also executes redis-cli commands for cache management and SQL queries for database session control.
[REMOTE_CODE_EXECUTION]: Remediation playbooks involve the use of npm and pip to install or upgrade dependencies, which fetches and executes code from remote registries. The skill also generates and applies code fixes directly to the repository via file edit tools.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its ingestion of untrusted data from the codebase, system logs, and external security advisories. It lacks explicit sanitization or boundary markers (e.g., delimiters) to separate this data from its instructions, which is critical given its broad capability inventory of shell and file system access.
[EXTERNAL_DOWNLOADS]: Playbooks reference external platforms like LaunchDarkly and Unleash for feature flag operations and utilize public package registries for dependency updates.

incident-remediating