The Agent Skills Directory

[SAFE]: The skill's primary logic is contained within local Python scripts (scripts/calendar_fragmentation.py and scripts/session_log_analyzer.py). These scripts perform mathematical analysis on user data using only the Python standard library and do not request network access, perform privileged system operations, or interact with external servers.
[COMMAND_EXECUTION]: The instructions direct the agent to execute the included Python scripts and perform basic file system operations such as directory creation (mkdir) to manage analysis reports. These operations are transparently defined and aligned with the skill's stated productivity purpose.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (calendar exports and session logs) which presents a potential attack surface for indirect prompt injection.
Ingestion points: External JSON files provided via the --input argument to the Python scripts.
Boundary markers: The scripts utilize standard JSON parsing; there are no explicit prompt isolation markers defined for the data when it is subsequently processed by the agent.
Capability inventory: The scripts are restricted to local file reading and writing to standard output; they do not have network access or dynamic execution capabilities (e.g., eval/exec).
Sanitization: The scripts perform validation on the JSON structure and verify the presence of required data fields (e.g., date, start, end) during analysis.

deep-work