The Agent Skills Directory

[SAFE]: The skill manages organizational memory using two layers of local markdown files (decisions.md and raw transcripts). All data operations are local to the memory/board-meetings/ directory.
[SAFE]: The Python script scripts/decision_tracker.py uses only Python standard libraries (argparse, os, re, sys, datetime, pathlib, typing). It does not import any network-capable modules like requests or urllib, nor does it attempt to execute arbitrary shell commands.
[SAFE]: There are no signs of prompt injection or obfuscation. The 'DO_NOT_RESURFACE' mechanism is a logical constraint for the agent to follow during future deliberations, rather than an attempt to bypass system safety guidelines.
[SAFE]: No external dependencies are downloaded or executed. The skill includes a clear template for manual data entry and a script for local parsing and reporting.

decision-logger