The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from outline/taxonomy.yml and GOAL.md. Malicious content within these files could influence the generated chapter skeleton, potentially affecting downstream agent tasks.
[DYNAMIC_EXECUTION]: The scripts/run.py script dynamically modifies the Python search path using sys.path.insert(0, str(repo_root)) to load the local tooling package. This is a common pattern in monorepos but involves runtime modification of the execution environment.
[COMMAND_EXECUTION]: The skill instructions define a workflow involving the execution of the local Python script scripts/run.py using uv run. This script performs file system operations including reading inputs, creating backups, and writing YAML output.

chapter-skeleton