The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from CDN metrics (URLs, paths) and error logs (log messages, body content) which are processed and interpolated into performance reports.
Ingestion points: Untrusted data enters the agent context through the MONITOR_TOP_PATHS, SEARCH_LOGS, and GET_LOG_DETAILS tools as described in SKILL.md and tools-reference.md.
Boundary markers: The skill does not explicitly define delimiters or warnings to ignore instructions embedded within the processed logs or URL paths.
Capability inventory: The skill is limited to read operations (monitoring/logging) and report generation. It does not possess capabilities for file writing, arbitrary command execution, or non-whitelisted network requests.
Sanitization: There is no mention of sanitization or filtering for potentially malicious strings within the ingested log data or URL paths before they are included in the markdown report output.
[DATA_EXFILTRATION]: The skill provides the ability to query and group logs by potentially sensitive fields.
Sensitive data access: The GET_LOG_DETAILS tool allows grouping and retrieving logs by fields such as userEmail and body (log message content), which could expose Personally Identifiable Information (PII) if present in the logs.
Context: Access to these logs is restricted to the site owner's infrastructure via the provided MCP tools and is consistent with the primary purpose of performing a performance and error audit.

deco-performance-audit