The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its automated research capabilities.
Ingestion points: The researcher agent (agents/researcher.md) uses the WebFetch tool to retrieve content from external websites during competitive and audience analysis.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions for the content retrieved from the web.
Capability inventory: The design-engineer (agents/design-engineer.md) and qa-lead (agents/qa-lead.md) agents have access to the Bash tool and file system access via Write, which could be exploited if malicious instructions from the web are processed as authoritative.
Sanitization: There is no evidence of sanitization or filtering of external data before it is passed to other agents in the workflow.
[EXTERNAL_DOWNLOADS]: The SKILL.md file contains instructions for an optional command npx designteam create. This command downloads and executes a package from the NPM registry at runtime.
[COMMAND_EXECUTION]: Multiple agents within the skill are granted powerful shell access.
The design-engineer (agents/design-engineer.md) and qa-lead (agents/qa-lead.md) are both provided with the Bash tool to implement and test code components, granting them significant access to the local execution environment.
The SKILL.md instructions include a setup step requiring the user to manually copy agent and command definitions into the project's configuration directory.
[DATA_EXFILTRATION]: The npx designteam create command is intended to transmit local role configurations to an external web service (designteam.app) to generate shareable links. This is a functional data transmission to the vendor's platform.

design-team