The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository (READMEs, source code) to generate tour descriptions. A malicious file could contain instructions designed to influence the agent's output.
Ingestion points: scripts/generate_from_docs.py reads README.md, CONTRIBUTING.md, and other documentation. The agent reads source files to fill in tour steps.
Boundary markers: Absent. The agent is instructed to fill in every [TODO: ...] without specific delimiters or warnings against embedded instructions.
Capability inventory: The skill executes local Python scripts for validation and generation. It has file read/write access to the repository.
Sanitization: Absent.
[COMMAND_EXECUTION]: The skill executes local Python scripts (validate_tour.py and generate_from_docs.py) to automate tour generation and validation. These scripts are bundled with the skill and use standard libraries.
[EXTERNAL_DOWNLOADS]: The skill fetches example CodeTour files from official GitHub repositories to provide the agent with real-world context and narrative structures.

code-tour