The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script, validate_factory_files.py, to perform validation on local repository files.
Evidence: The skill instructs the agent to run python3 "{{skill_dir}}/scripts/validate_factory_files.py".
[EXTERNAL_DOWNLOADS]: The skill fetches JSON schemas from the official Warp developer API to ensure local configurations align with current server requirements.
Evidence: curl -s https://app.warp.dev/api/v1/factory-files/schemas is used to retrieve the supported schema versions.
[DATA_EXFILTRATION]: The validation script reads local factory configuration files and submits their content to the vendor's validation endpoint via a POST request.
Evidence: validate_factory_files.py collects file content and sends it to https://app.warp.dev/api/v1/factory-files/validate. The script includes security controls such as _leaves_factory_root to prevent reading files outside the designated directory via symlinks.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes and validates user-provided repository files (YAML and Markdown) which contain agent instructions and rubrics.
Ingestion points: Local files under the factory-root directory (e.g., factory.yaml, agents/*.md).
Boundary markers: None explicitly enforced by the validation script beyond schema compliance.
Capability inventory: The skill has file-read capabilities and network-write capabilities (sending data to the vendor API).
Sanitization: The validation script verifies the structure against a schema but does not sanitize the natural language content within the files.

factory-files