The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by writing user-supplied data to a company-context.md file, which is subsequently read by all other agent roles as authoritative context.
Ingestion points: The /cs:setup command (SKILL.md) collects seven points of business data from the user.
Boundary markers: Absent; there are no instructions for the agents to treat the content of company-context.md as untrusted data or to disregard potential instructions embedded within it.
Capability inventory: The skill utilizes 25 CLI-based Python tools and has the ability to write files to the project root.
Sanitization: Absent; user responses are written directly into a structured markdown file without validation or escaping.
[DATA_EXFILTRATION]: The skill encourages the storage of sensitive corporate intelligence in plaintext files within the repository.
The /cs:setup workflow explicitly requests high-value financial data such as ARR (Annual Recurring Revenue), MRR, and runway metrics. Storing this data in company-context.md in the project root creates a data exposure risk for any collaborator or process with access to the filesystem or version control.
[COMMAND_EXECUTION]: The skill relies on 25 Python-based tools described as "CLI-first."
While the skill claims these are dependency-free, the use of CLI tools to generate JSON output suggests a broad capability for the agent to execute system commands and perform local environment operations during strategic simulations and board meetings.

c-level-advisor