The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to collect and store highly sensitive business data including revenue figures, burn rates, runway length, and internal strategic risks. While this data is stored locally in ~/.claude/company-context.md, the solicitation and consolidation of such sensitive information into a single known file path increases the impact of potential future data theft.
[COMMAND_EXECUTION]: The skill instructions include the execution of a shell command (ln -sf) to create a symbolic link between a local vault and the agent's configuration directory.
[INDIRECT_PROMPT_INJECTION]: This skill serves as a data ingestion point where untrusted user input is collected and stored to be used as authoritative context by other agents, creating a surface for indirect prompt injection.
Ingestion points: Responses provided by the user to the 12 interview questions defined in SKILL.md.
Boundary markers: Absent. The output template uses standard Markdown headers but lacks delimiters or instructions to prevent the agent from interpreting embedded malicious commands in the collected data.
Capability inventory: The skill has the capability to write files to the local file system (~/.claude/company-context.md) and perform shell operations.
Sanitization: Absent. The instructions explicitly direct the agent to quote the founder's own words verbatim, which encourages the direct interpolation of untrusted strings into the persistent context file.

onboard