The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external data sources such as interview transcripts, support tickets, and GitHub pull requests.
Ingestion points: Data is ingested via workflow steps in SKILL.md and interview guides in references/source-interview-guide.md.
Boundary markers: The instructions do not mandate specific technical delimiters for untrusted data, but they require a human-in-the-loop approval process.
Capability inventory: No code execution, file writing, or network operations are present in the skill.
Sanitization: The skill includes a 'Publication gate' that requires human review of technical accuracy and a specific 'Approval track' for redacting sensitive information.
[SAFE]: The skill contains no executable code, package dependencies, or shell commands. It functions strictly as a high-level instruction set for the AI agent.
[SAFE]: The skill references established industry resources and case studies from well-known technology organizations (e.g., Temporal, Honeycomb, PostHog, Vercel) to provide templates and examples.
[SAFE]: Proactive security instructions are provided in references/approval-and-anonymization.md, directing the agent to redact sensitive information like cluster sizes, internal service codenames, and ticket IDs to prevent accidental data exposure.

developer-case-study