The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests founder interview responses and stores them in a persistent context file without sanitization or explicit boundary markers. * Ingestion points: User responses provided during the /cs:setup and /cs:update interview flows in SKILL.md. * Boundary markers: The skill uses markdown headers defined in templates/company-context-template.md to structure data, but does not provide instructions to the agent to treat this content as untrusted or ignore embedded instructions. * Capability inventory: The skill instructions in SKILL.md direct the agent to write to the local filesystem at ~/.claude/company-context.md. * Sanitization: There is no evidence of filtering, escaping, or validation of input strings before they are written to the context file.
[DATA_EXFILTRATION]: The skill instructions direct the agent to capture and store highly sensitive business information in a persistent local file. * Sensitive Data: Captures revenue ranges, runway length, headcount, internal team conflicts, and founder vulnerabilities. * Storage Path: Stores this information at ~/.claude/company-context.md, creating a persistent target containing sensitive business intelligence.

cs-onboard