The Agent Skills Directory

[SAFE]: The skill operates entirely on local files within the project directory to manage meeting transcripts and decisions. Analysis of the scripts and instructions revealed no malicious patterns, obfuscation, or unauthorized data access.
[COMMAND_EXECUTION]: The provided Python script (scripts/decision_tracker.py) is used for parsing local markdown files and generating reports. It relies solely on the Python standard library and does not perform any dangerous operations like executing arbitrary shell commands or downloading external content.
[DATA_EXPOSURE]: No evidence of sensitive data exfiltration was found. The skill processes internal meeting data but lacks network capabilities or access to sensitive system paths (e.g., SSH keys, AWS credentials).
[INDIRECT_PROMPT_INJECTION]: While the skill automatically loads past decisions into the context of future meetings, the risk of indirect prompt injection is mitigated by a mandatory human-in-the-loop workflow.
Ingestion points: Past decisions are loaded from memory/board-meetings/decisions.md.
Boundary markers: The system uses structured Markdown headers and specific field labels defined in templates/decision-entry.md to delimit data.
Capability inventory: The skill uses scripts/decision_tracker.py to parse text and detect conflicts; no dangerous execution or network capabilities are present.
Sanitization: All decisions and synthesis must be explicitly approved by the founder before being committed to the persistent memory layer.

decision-logger