The Agent Skills Directory

Write-Safety Protocols: The skill implements a strong security posture regarding data modification. It requires a mandatory review process where the agent must display a tabular comparison of current versus proposed values and obtain explicit user consent before committing changes to the CRM. This directly mitigates risks associated with accidental data corruption or unauthorized updates.
User Context Verification: The skill requires an initial preflight check via a 'user-context' module to establish authorized scope and source mapping. This ensures that the agent operates within the user's defined permissions and context before accessing sensitive CRM records.
Controlled Data Operations: Guidance is provided to use specific, non-deprecated tools for searching and retrieving records, which reduces the surface area for unexpected API behaviors. It also enforces limits on batch sizes (max 10 objects) to prevent high-impact bulk changes.
Data Handling and Attribution: The instructions emphasize providing clickable URLs with UTM parameters for auditing purposes and require the agent to disclose whether analysis is based on sampled data, promoting transparency and traceability in data handling.

hubspot