The Agent Skills Directory

[SAFE]: The skill demonstrates secure design principles by utilizing a step-by-step wizard format that prevents automated or silent file modifications. All data operations targeting the local /foundation/brain.md file require an explicit 'yes' from the user after viewing a summary of changes.
[INDIRECT_PROMPT_INJECTION]: The skill has the capability to ingest data from external platforms (such as Slack, HubSpot, and Notion) via Model Context Protocol (MCP) servers. While this data could theoretically contain malicious instructions, the risk is effectively mitigated by the skill's 'Operating Rules' which treat all pulled facts as unconfirmed drafts and enforce a manual review process before the data enters the agent's long-term context.
Ingestion points: External data provided by Notion, Atlassian, HubSpot, Fireflies, Gong, SimilarWeb, Intercom, Google Drive, and Slack MCP servers.
Boundary markers: The skill uses a confirmation gate ('Look good? (yes / edit / skip)') for every section update.
Capability inventory: Writing to the local file system (/foundation/brain.md and .brain-draft.md).
Sanitization: Implements logic to reject generic or vague inputs and tags all external data as 'draft' until verified by the human user.
[EXTERNAL_DOWNLOADS]: The skill configuration in .mcp.json references several well-known and trusted technology services (including Google, Slack, Notion, and HubSpot) to fetch organizational context. These connections are standard for the skill's stated purpose of aggregating marketing data and do not involve executing untrusted code.

product-marketing-context