The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes application logs and configuration files to diagnose integration issues, presenting a potential surface for indirect prompt injection. This risk is addressed through explicit redaction requirements and structured data extraction.
Ingestion points: Step 1 in SKILL.md utilizes the Read and Grep tools to collect evidence from relevant configuration and log files.
Boundary markers: The references/error-triage-playbook.md file defines a strict evidence boundary, explicitly excluding credentials, API keys, prompts, and raw customer data from analysis.
Capability inventory: The skill utilizes Read, Grep, and Bash (restricted to curl commands for diagnostic probes to official Mistral AI endpoints).
Sanitization: The instructions mandate the use of jq for structured extraction of JSON fields and require a redacted final report format that omits sensitive identifiers.

mistral-common-errors