The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/run.py dynamically computes the repository root by traversing parent directories searching for a specific file (AGENTS.md) and then inserts this path into sys.path. This allows the script to import modules from a variable filesystem location. While common in some development environments, dynamic loading from computed paths can be risky if the environment is not strictly controlled.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from sources/index.jsonl and sources/provenance.jsonl to generate tutorial specifications. This creates a surface for indirect prompt injection where malicious instructions embedded in the ingested source corpus could influence the agent's output.
Ingestion points: sources/index.jsonl and sources/provenance.jsonl read via build_source_tutorial_spec in scripts/run.py.
Boundary markers: None identified in the provided code to delimit untrusted data from instructions.
Capability inventory: The skill has file-writing capabilities via atomic_write_text in scripts/run.py.
Sanitization: No explicit sanitization or filtering of the source data is visible in the skill wrapper script.

source-tutorial-spec