The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch data from external research APIs (like Perplexity) and distill the findings into hypothesis sets. This process ingests untrusted data from the web which could contain malicious instructions designed to influence the agent's output or subsequent actions.
Ingestion points: External research API responses gathered in Step 2 and Step 4.
Boundary markers: Absent. The instructions do not specify the use of delimiters or warnings to ignore embedded instructions in the research data.
Capability inventory: The skill performs file system write operations to the claude-code-gtm/context/ directory and performs network operations via the chosen API provider.
Sanitization: Absent. There is no mention of escaping or validating the content retrieved from external APIs before saving it to files or using it in further processing.
[COMMAND_EXECUTION]: The workflow requires the agent to perform network requests and file system operations.
Network operations: The agent is expected to interact with third-party APIs (Step 0 and Step 2) to perform deep research.
File access: The agent reads from claude-code-gtm/context/{company}_context.md and writes research outputs to several files within the claude-code-gtm/context/{vertical-slug}/ directory.

market-research