The Agent Skills Directory

[SAFE]: The skill serves as a best-practices reference for Cognite Data Fusion (CDF) Data Modeling. It includes a robust implementation of a QueuedTaskRunner (semaphore) to manage concurrency and avoid 429 Too Many Requests errors when calling external APIs.
[SAFE]: All external documentation references point to the official vendor domain (docs.cognite.com), which is a trusted resource associated with the skill's authoring organization.
[SAFE]: The skill demonstrates a proactive security posture by including a 'Hard gate' section that explicitly restricts the use of Large Language Models (LLMs) over query results. It recommends strict limits (5-50 calls), mandatory caching, and ensuring operations are user-initiated to mitigate risks associated with processing potentially untrusted external data.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of data from external CDF APIs. While this creates a surface for indirect prompt injection, the skill specifically provides remediation guidance (volume caps and caching) to manage how an agent processes this data.
Ingestion points: CDF instances.list, instances.query, and instances.search API patterns in SKILL.md.
Boundary markers: The skill mandates a 'Hard gate' policy for LLM calls over results.
Capability inventory: The agent has access to Read, Glob, Grep, Edit, and Write tools.
Sanitization: Guidance provided in the 'Hard gate' section advises on limiting and caching completions to prevent uncontrolled processing of API output.

dm-limits-and-best-practices