The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructions direct the agent to access sensitive credential data stored on the local file system at ~/.build-host/credentials.json. Access to such files is a high-risk operation as it exposes authentication secrets to the agent context.
[COMMAND_EXECUTION]: The operating procedure describes the use of APIs to perform deployments and runtime operations, which involves executing code and managing services on external infrastructure based on user-provided scripts.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data from project directories and git repositories to determine deployment logic.
Ingestion points: Project directory contents, git repository files, and target branches.
Boundary markers: None identified. There are no instructions to use delimiters or warnings to ignore potentially malicious instructions embedded within the user's project files.
Capability inventory: Reading local filesystem credentials, interacting with the build.host API, and executing deployment/operation tasks.
Sanitization: The instructions do not specify any validation or sanitization of the user-provided project code before it is processed and deployed.

deploy-cron