The Agent Skills Directory

Analysis of External Content: The skill processes call transcripts and summaries through the meeting_notes tool. This is an indirect data ingestion point where instructions embedded in transcripts could theoretically influence the agent. The skill mitigates this by requiring an 'evidence-first' approach, utilizing a strict analysis rubric in references/rubric.md, and limiting quotes to 25 words. These boundary markers and sanitization strategies ensure the agent remains focused on objective behavior analysis.
Internal Skill Dependencies: The skill workflow involves a preflight invocation of a local helper skill (../user-context/SKILL.md). This modular design facilitates consistent user context and source resolution across a suite of tools without introducing external security risks or remote execution patterns.
Data Privacy and Integrity: The instructions include explicit rules against exposing raw connector IDs in the final output and require absolute date ranges to be calculated before querying tools. These practices protect internal metadata and ensure that the evidence provided to the user is grounded in concrete, verifiable time windows.
Constraint-Based Reasoning: The skill defines a clear capability inventory, limited primarily to searching and fetching meeting notes for analysis. It does not possess capabilities for unauthorized network operations or file writes, adhering to the principle of least privilege.

review-rep-call-trends