The Agent Skills Directory

[PROMPT_INJECTION]: The skill mandates 'AUTONOMOUS MODE' and explicitly instructs the agent: 'Do NOT ask questions. Do NOT pause for confirmation.' This configuration attempts to bypass the essential safety filter of human review and user intervention during the execution of potentially destructive file modifications.\n- [DATA_EXFILTRATION]: The skill accesses highly sensitive file path patterns as part of its discovery phase, including '.env', '.env.*', 'vault/', 'secrets/', and searches for hardcoded API keys or tokens within the source code. This exposes sensitive environment credentials to the agent's context.\n- [COMMAND_EXECUTION]: The skill performs unauthorized file system operations by writing execution telemetry to '~/.claude/projects/skill-telemetry.md', which is located in the user's home directory outside of the intended project scope.\n- [PROMPT_INJECTION]: The skill is highly vulnerable to indirect prompt injection due to its core functionality of reading and acting upon untrusted project files.\n
Ingestion points: The skill scans all project configuration files across CI/CD, containerization (Docker), orchestration (Kubernetes), and Infrastructure-as-Code (Terraform) domains.\n
Boundary markers: No delimiters or instructions are used to ensure the agent ignores malicious directives that might be embedded within the scanned configuration files.\n
Capability inventory: The skill has the capability to modify project infrastructure files, create new files in the project root, and write to the user's home directory.\n
Sanitization: There is no evidence of sanitization or content validation for the data ingested from external files before it is used to generate reports or perform remediations.

devops