The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the execution of local bash and PowerShell scripts to manage the lifecycle of Docker containers. It explicitly instructs the AI agent to attempt to use environment-specific privilege escalation mechanisms, such as sudo, if Docker commands fail due to permission errors.
[EXTERNAL_DOWNLOADS]: The included scripts are configured to fetch the specmatic/enterprise:latest Docker image from the vendor's repository if no local Specmatic image is detected.
[DATA_EXFILTRATION]: The execution logic mounts the user's home directory subdirectory ~/.specmatic (intended for license storage) and the current working directory into the Docker container, potentially exposing local configuration and sensitive project files to the containerized environment.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it processes external OpenAPI specification files. A malicious specification could contain instructions in field descriptions designed to influence the agent's logic during the analysis cycle.
Ingestion points: OpenAPI specification files (SKILL.md).
Boundary markers: No boundary markers or specific delimiters are defined to separate the specification content from the agent's instruction context.
Capability inventory: The skill executes shell scripts, performs Docker operations, and writes to the local filesystem (logs and updated specifications).
Sanitization: No sanitization or validation of the specification's natural language fields is performed before the agent processes and acts upon the data.

fix-openapi-spec