8288bc0b77
Inject a human-readable summary of the active SecurityPolicy into the system prompt Safety section. LLM sees allowed commands, forbidden paths, autonomy level, and rate limits. Supersedes #3968. Closes #2404