The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses the WebFetch tool to download content from competitor URLs discovered in the local .design/branding/{brand}/BRIEF.md file. This content is then provided as context to the research agent.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8c). Content fetched from external websites is inlined into the gsp-brand-researcher agent's prompt without visible sanitization or robust boundary markers. An attacker who controls a competitor website could embed malicious instructions to manipulate the research output or the agent's actions.
Ingestion points: Competitor websites are fetched via WebFetch in Step 2.5 and the resulting content is inlined into the agent prompt in Step 3.
Boundary markers: The skill does not define specific delimiters or instructions for the agent to ignore potential commands embedded in the fetched web content.
Capability inventory: The agent has access to Write and Bash tools, meaning an injection could potentially lead to unauthorized file modifications or command execution.
Sanitization: There is no evidence of content filtering or sanitization performed on the fetched external data before it is processed by the agent.

gsp-brand-research