The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides logic to ingest and process external data from Langfuse traces (observations and generations) to generate cost reports and budget alerts. If trace data contains malicious instructions, it could potentially influence the agent when it reads the generated reports.
Ingestion points: Data is retrieved from the Langfuse API via langfuse.api.traces.list, langfuse.api.observations.list, and langfuse.fetchGenerations (found in SKILL.md and references/implementation.md).
Boundary markers: The provided code snippets do not include explicit delimiters or 'ignore embedded instructions' warnings when processing trace data.
Capability inventory: The skill utilizes Read, Write, and Edit tools, and the implementation examples include fetch for sending data to Slack webhooks.
Sanitization: No sanitization or filtering of external content is present in the provided implementation examples.
[SAFE]: The skill uses well-known and official libraries for Langfuse (@langfuse/client, @langfuse/openai, @langfuse/tracing) and OpenAI integration. It follows best practices by recommending the use of environment variables for sensitive credentials like SLACK_WEBHOOK_URL.

langfuse-cost-tuning