The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads content from a target repository (specifically README files) using scripts/inspect_repo.py to extract titles and descriptions for launch materials. Malicious instructions embedded in these files could attempt to influence the agent's behavior during the generation of the launch pack.
Ingestion points: scripts/inspect_repo.py reads file content from the provided repository path.
Boundary markers: None identified in the script logic for the extracted text.
Capability inventory: File reading in inspect_repo.py, file writing and directory creation in render.py and render_card.py, and subprocess execution in render.py and the test suite.
Sanitization: scripts/render_card.py employs html.escape for SVG rendering, which mitigates XSS-style injection in the image assets, but the agent processes the raw extracted text before this stage.
[COMMAND_EXECUTION]: scripts/render.py utilizes subprocess.check_call to orchestrate the rendering process by calling render_card.py. The command is constructed using the current Python executable and absolute paths to the skill's own internal scripts, representing a low-risk internal orchestration pattern.
[DYNAMIC_EXECUTION]: scripts/render_card.py uses __import__('os') to dynamically check for the presence of the EDGE_SKIP_PNG environment variable. While this is a dynamic module load, it is restricted to a standard library module.

repo-to-launch