The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses sensitive information in docs/LAUNCHES.md and 4-Mini-Launch.md, which contain data on signups, activated users, and payments. It also reads the Product Offer and Customer Persona to form hypotheses.
[DATA_EXFILTRATION]: The workflow requires the agent to use "web search and any connected research tools" to find competitor pricing and benchmarks. Business details from the sensitive input files could potentially be exposed to search engines or third-party research tools during this process.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the public web, such as competitor pricing pages. This data is used to propose business hypotheses and critique the user's choices. Maliciously crafted content on these external sites could attempt to influence the agent's strategic advice or behavior.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The agent reads internal files (Product Offer, Customer Persona, Launch Log) and external data from web search results.
Boundary markers: The instructions do not define specific delimiters or instructions to prevent the agent from following directives found in external research results.
Capability inventory: The skill has the ability to read project files, perform web searches, and overwrite the BONUS-Business-Strategy-Deep-Dive.md file.
Sanitization: There is no mention of sanitizing or validating content retrieved from the web before incorporating it into the strategy document.

studio-define-business-strategy