The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface. It collects untrusted user input (interview responses) and stores it in a persistent location (~/.claude/company-context.md) specifically designed to be read by other C-suite advisor skills.
Ingestion points: Data is collected through user dialogue during the /cs:setup and /cs:update workflows described in SKILL.md.
Boundary markers: Absent. The templates/company-context-template.md does not define clear boundaries or data/instruction separators to ensure downstream skills treat the content as untrusted data.
Capability inventory: The agent must use file-writing capabilities to generate the persistent context file.
Sanitization: Absent. Malicious instructions provided by a user during the interview are written directly to the context file, potentially compromising other skills that ingest this file.
[COMMAND_EXECUTION]: The skill directs the agent to write to a configuration directory in the user's home folder (~/.claude/). While this is central to the skill's functionality as a persistent context provider, modifying files in home-directory configuration paths is a sensitive operation that exceeds typical project-level file access.

cs-onboard