The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes data from historical session logs and lessons to generate new executable skill files, creating a surface for indirect prompt injection. * Ingestion points: Historical data is read from .great_cto/logs/session-*-end.md and .great_cto/lessons.md in Step 1 and Step 2. * Boundary markers: No explicit delimiters or instructions are provided to the sub-agent to distinguish between trusted instructions and untrusted data in the logs. * Capability inventory: The skill has the capability to Write new files to the skills/ directory and execute Bash commands. * Sanitization: No evidence of sanitization or validation of the extracted content before it is written into new draft skill files.
[COMMAND_EXECUTION]: In Step 4, the skill utilizes node -e to execute a dynamically constructed Javascript string to write metadata to a file. While the logic is simple, it involves interpolating shell-derived variables directly into the executed script.

crystallize