The Agent Skills Directory

[SAFE]: The skill is designed to improve the operational safety and reliability of AI agents. It provides structured playbooks for reviewing observability, optimization loops, and autonomous controllers.
[SAFE]: No hardcoded credentials, malicious network operations, or unauthorized file access patterns were identified across the 24 files.
[SAFE]: The skill includes explicit security recommendations, such as using privacy filters to scrub captured content before persistence and requiring human review for agent-authored changes.
[PROMPT_INJECTION]: The skill is designed to process external agent traces, which constitutes a potential surface for indirect prompt injection. This is inherent to its primary purpose as an auditing tool and is addressed through explicit guidance on sanitization and privacy filtering.
Ingestion points: Processes external agent traces and session logs (SKILL.md, playbooks).
Boundary markers: Relies on structural parsing of trace data; no specific prompt delimiters are prescribed in templates.
Capability inventory: Writing audit reports and findings ledgers to the local filesystem (docs/audits/); spawning sub-agents for specialized lens analysis.
Sanitization: Playbooks explicitly recommend implementing privacy filters and compaction policies to scrub captured content before persistence.

agent-ops