The Agent Skills Directory

[SAFE]: The skill consists entirely of natural language instructions designed to enforce a strict reasoning discipline (evidence-chain) for chat interactions. It does not include any executable scripts, binary files, or configuration that triggers command execution.
[SAFE]: No patterns of prompt injection, data exfiltration, or obfuscation were detected. The instruction for the agent to cite sources (including web links) is a transparency requirement for the agent's output, not a network operation performed by the skill itself.
[SAFE]: References to external files like 'evidence_terminal.py' and 'harness/policy.yaml' are mentions of the existing environment's validation infrastructure and do not involve downloading or executing untrusted code.

graph-mode