The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a notable attack surface for indirect prompt injection due to its role in processing external, potentially attacker-controlled data using high-privilege automation tools.
Ingestion points: The skill is designed to read and analyze external data sources such as application logs (kubectl logs), git commit history (git log), and CI/CD environment variables.
Capability inventory: The skill possesses extensive operational capabilities including cluster management (kubectl), repository modification (gh, git), infrastructure provisioning (terraform), and container operations (docker).
Boundary markers: There are no explicit instructions or structural delimiters (e.g., XML tags or specific 'ignore' warnings) provided to the agent to distinguish between internal instructions and ingested external data.
Sanitization: The provided automation templates and shell scripts do not implement rigorous input validation or escaping for external data interpolated into command-line arguments.
[COMMAND_EXECUTION]: The skill implementation relies heavily on shell script generation and execution as part of its primary DevOps automation purpose.
Evidence: Multiple files (references/incident-response.md, references/platform-engineering.md, references/release-automation.md) contain bash and python scripts that execute system-level commands to manage infrastructure and deployments.
[DATA_EXPOSURE]: The skill provides automated forensics and incident response procedures that involve the collection of sensitive technical metadata.
Evidence: The collect-evidence.sh script in references/incident-response.md automates the extraction of pod logs, database activity snapshots (pg_stat_activity), and detailed Kubernetes resource configurations to local storage.

devops-engineer