The Agent Skills Directory

[SAFE]: The skill operates entirely on local workspace files to generate survey prose. It reads input from files such as DECISIONS.md, outline/outline.yml, and outline/writer_context_packs.jsonl, and writes the resulting sections to the sections/ directory.
[SAFE]: The Python execution script scripts/run.py uses standard libraries for text processing and file management. No suspicious network operations, credential harvesting, or dynamic code execution patterns were detected.
[SAFE]: Extensive regex-based filtering (e.g., _neutralize_author_voice, _WRITER_UNSAFE_SUPPORT_RE) is implemented to maintain prose quality by removing filler text and authorial artifacts, with no evidence of malicious intent or obfuscation.
[SAFE]: The skill follows secure file writing practices, including a bootstrap mechanism that requires human approval (Approve C2 in DECISIONS.md) and uses atomic writes for file updates.

subsection-writer