The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines automated auditing and triage workflows that ingest potentially attacker-controlled external data.
Ingestion points: Procedures in energy-auditor.md, memory-auditor.md, and swift-performance-analyzer.md ingest codebase content. The hang-diagnostics.md and metrickit-ref.md files describe triaging .ips crash logs, Sentry reports, and MetricKit diagnostic payloads.
Boundary markers: The auditor templates and triage instructions do not implement delimiters or 'ignore embedded instructions' warnings for the data being analyzed.
Capability inventory: The skill utilizes analysis tools including Glob, Grep, Read, xctrace, xcprof, and xcsym, which allow for broad filesystem and process metadata access.
Sanitization: The instructions do not define requirements for sanitizing, escaping, or filtering content from the analyzed logs or reports before processing by the agent.

axiom-performance