The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes structured data (JSON) and raw HTML fragments provided by the user to generate artifacts. This ingestion of untrusted data represents an attack surface for indirect prompt injection if the output is later processed by another agent. However, the skill includes explicit documentation on safe handling and implements automatic sanitization for non-HTML fields.
Ingestion points: The scripts/build.py script ingests data via --spec files and --set command-line overrides.
Boundary markers: The skill does not implement specific boundary markers in the generated HTML for data interpolation.
Capability inventory: The skill has the capability to read from and write to the local filesystem during the build process.
Sanitization: scripts/composer.py provides an esc() function that performs standard HTML escaping using the Python html library.
[COMMAND_EXECUTION]: The skill's test suite, specifically tests/test_smoke.py, uses subprocess.run to perform integration tests on the command-line interface. This execution is confined to the test environment and is a standard practice for verifying the functional correctness of the build tool.

composing-html