The Agent Skills Directory

[PROMPT_INJECTION]: No malicious injection patterns detected. The skill contains standard instructional language and 'Anti-Hallucination Rules' that are defensive in nature and intended to ensure accuracy in business deliverables.
[DATA_EXPOSURE]: No hardcoded credentials, API keys, or access to sensitive system file paths detected. The skill processes user-provided business notes, which is its intended function.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns, package installations, or dynamic code execution detected. The skill operates entirely through text-based prompts and references within its own folder structure.
[OBFUSCATION]: No obfuscated URLs, Base64-encoded commands, zero-width characters, or homoglyph attacks detected in the markdown or reference files.
[PRIVILEGE_ESCALATION]: No privilege escalation commands (sudo, chmod, etc.) detected.
[PERSISTENCE]: No persistence mechanisms (cron jobs, startup scripts, registry modifications) detected.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data (CSM notes, customer dumps) but lacks dangerous capabilities (subprocess calls, file-write, network ops) that could be exploited via injection. It also includes explicit 'Untrusted content' warnings to the agent to treat data as data, not instructions.
[DYNAMIC_CONTEXT_INJECTION]: No use of the dynamic context injection syntax (!command) detected.

onboarding-plan