The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions direct the agent to ingest external, untrusted data to populate the documentation templates.
Ingestion points: The workflow (Step 1 in SKILL.md) requires the agent to read project specifications, READMEs, and source code to identify technical limits, constants, and logic for inclusion in the manual.
Boundary markers: The instructions do not specify the use of delimiters or clear boundary markers when interpolating this untrusted data into the HTML templates.
Capability inventory: The skill utilizes shell scripts (render.sh), Python scripts (check.py), and file system operations (mkdir, cp) to manage the document generation lifecycle.
Sanitization: There are no explicit instructions for sanitizing or escaping the data extracted from source files before it is rendered into the final HTML/PDF output.
[COMMAND_EXECUTION]: The skill relies on local script execution to perform its primary functions.
Evidence: SKILL.md instructs the agent to execute sh scripts/render.sh for PDF generation and uv run ... python scripts/check.py for visual validation. These scripts are part of the skill's distributed package and are required for its intended purpose.
[DYNAMIC_EXECUTION]: The rendering process involves dynamic execution of template code.
Evidence: The skill's HTML templates (assets/template.html and examples) contain embedded JavaScript used for 'fit checks' (measuring DOM elements to prevent overflow). This code is executed by a headless Chrome instance controlled by render.sh.
[EXTERNAL_DOWNLOADS]: The skill requires external dependencies from a well-known registry.
Evidence: The validation script (check.py) requires the pymupdf library, which the instructions suggest installing via uv or standard Python package managers. PyPI is a recognized and established service for software dependencies.

nasa-handbook