The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external files in the workspace to build planning substrate, creating a vector for malicious instructions to reach the agent's context during later steps.
Ingestion points: scripts/run.py reads content from outline/chapter_skeleton.yml, outline/section_bindings.jsonl, and GOAL.md provided in the workspace.
Capability inventory: The script performs file system operations, specifically writing the assembled planning data to outline/section_briefs.jsonl via the _write_jsonl function.
Boundary markers: The output JSONL does not utilize explicit boundary markers or instructions to isolate the data from the agent's control flow.
Sanitization: No sanitization or filtering is applied to the text content extracted from the input files before it is interpolated into the generated brief artifacts.

section-briefs