The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by capturing raw user input via the /btw {suggestion} command and storing it in .bkit/btw-suggestions.json. This untrusted data is later retrieved and processed by the /btw analyze and /btw promote commands to perform clustering, name generation, and as a seed for new skill creation.
Ingestion points: User input provided to the /btw command is saved to .bkit/btw-suggestions.json (SKILL.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the suggestion text are defined for the storage or analysis phases.
Capability inventory: The agent uses Read, Write, and Bash tools to manage these suggestions and initiate the skill-create pipeline.
Sanitization: There is no evidence of validation or sanitization of the suggestion content before it is used to influence the agent's logic in the analyze and promote steps.

btw