The Agent Skills Directory

[PROMPT_INJECTION]: No malicious override or bypass instructions detected. The 'Required Skills' and 'Orchestrator Definition' sections use instructional language to define operational flow, not to escape safety guardrails.
[DATA_EXFILTRATION]: No network calls or sensitive file path access (e.g., .ssh, .env) were found. The skill operates on structured data provided by the user and shared between spawned sub-agents.
[REMOTE_CODE_EXECUTION]: No remote code downloads or execution patterns (like curl | bash) are present.
[COMMAND_EXECUTION]: The skill uses spawn_agent which is a platform-native orchestration tool. No arbitrary shell command execution or privilege escalation was detected.
[OBFUSCATION]: The instructions are written in clear, plain-text markdown with no hidden characters, Base64 strings, or encoded URLs.
[DYNAMIC_CONTEXT_INJECTION]: There are no uses of the !command syntax that would execute shell commands at skill load time.

recipe-diagnose