The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local scripts to query and update a persistent knowledge base (the "brain").
Evidence: Skill executes python3 ../professional-brain/scripts/brain_query.py ./brain "churn" to retrieve existing context.
Evidence: Skill executes ../professional-brain/scripts/brain_write.py … --commit to save findings, decisions, and hypotheses.
[PROMPT_INJECTION]: The skill processes untrusted external data which creates a surface for indirect prompt injection.
Ingestion points: Untrusted data enters via "Churn reasons data" (exit survey results, CSM notes, support data) and "Customer segments" input fields.
Boundary markers: No explicit delimiters or instructions to ignore embedded instructions are used when processing the churn reasons or survey quotes.
Capability inventory: The skill has the capability to both read from and write to the local file system via the brain_query.py and brain_write.py scripts.
Sanitization: There is no evidence of input validation or sanitization of the customer-provided feedback before it is analyzed or recorded.

churn-analysis