The Agent Skills Directory

[SAFE]: The skill instructions define a process for summarizing chat progress and visual status updates. It explicitly mandates that the agent must not change code, start applications, or perform irreversible actions during the summary generation process ('Keep this turn read-only'). This instruction reduces the risk associated with Indirect Prompt Injection, as the agent is discouraged from acting on instructions found in the chat history or subagent reports it processes. No malicious patterns, obfuscation, or unauthorized data access were detected.

pstack-ketchup