The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it ingests data from external local files that may contain untrusted content.
Ingestion points: Step 1 involves reading content from PROJECTS.md and the sessions/ directory, which may include externally-sourced project descriptions or activity logs from varied sources.
Boundary markers: The instructions do not define delimiters or specific constraints to prevent the agent from accidentally following instructions that might be embedded within those files.
Capability inventory: The skill possesses the ability to execute shell commands via the gog CLI and perform file system writes to now.md, profile.md, and the assessments/ directory.
Sanitization: There is no evidence of sanitization or validation of the ingested file content before it is summarized or used to influence the agent's logic.
[COMMAND_EXECUTION]: The skill executes the gog command-line utility to automate the creation of events in Google Calendar. While the operation is performed after user confirmation of the assessment summary, it represents a direct shell command execution capability.

year-review