The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes untrusted data from external files and user input to generate reasoning outputs.
Ingestion points: The skill reads company profiles, win cases, and product information from claude-code-gtm/context/{company}_context.md and accepts direct user input regarding vertical markets.
Boundary markers: The instructions do not define specific delimiters or guardrails to prevent the agent from following instructions potentially embedded within the context files.
Capability inventory: The skill uses the agent's capabilities to read from and write to the local file system (specifically claude-code-gtm/context/ directory).
Sanitization: There is no evidence of content validation or sanitization for the markdown data retrieved from the context files before it is used for hypothesis generation.
[NO_CODE]: The skill consists entirely of markdown-based instructions and logic without including or invoking external scripts, executables, or code-based dependencies.

hypothesis-building