The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external observability data (traces, sessions, and signals) from Latitude APIs to generate automated findings and summaries. This represents a potential surface for indirect prompt injection if the processed data contains adversarial content designed to influence the agent's interpretation logic.
Ingestion points: Observability data fetched via queryAnalytics, listTraces, and other Latitude API operations in SKILL.md and data-sources.md.
Boundary markers: The instructions focus on data aggregation and interpretation but do not specify strict delimiters for the data interpolation in the final findings generation.
Capability inventory: The agent has the ability to write HTML files to the local file system and generate executable scripts (SKILL.md, refresh.md).
Sanitization: The provided HTML template includes an esc() function for data rendered in the UI, but the interpretation step relies on the LLM's internal reasoning.
[COMMAND_EXECUTION]: The skill provides templates in refresh.md for creating Shell, Python, and Node.js scripts to automate data fetching and report generation. These scripts are intended to be executed in the user's local environment or CI/CD pipelines as part of the dashboard functionality.
[PERSISTENCE]: The skill includes instructions in refresh.md for the agent to help users set up scheduled tasks via cron or GitHub Actions workflows to automate the dashboard refresh process. While this is a primary feature for maintaining up-to-date KPIs, it constitutes a persistence mechanism.

latitude-artifacts