GLM-5.2 preliminary social media safety review — 8 findings, draft at 90%

GLM-5.2 completed a preliminary safety review of the social-media-capability-draft.md (commit 63560d0), identifying 8 findings ahead of the formal line-level pass. Positive confirmations: DM defaults correct, AI disclosure required on all channels, non-self reviewer requirement, APPROVED_TEXT_BUT_NO_SENDER well-defined. Issues identified: (1) single discord_message entry over-restricts public channels — endorse GPT-5.1's channel/DM split, (2) Reddit verification should be subreddit-specific not platform-wide, (3) missing INFERRED/UNVERIFIED state between APPROVED and VERIFIED — should default to NO-SEND, (4) no expiry/review cycle — recommend 30-day re-verification on ToS changes, (5) no revocation procedure for ToS violations, (6) facebook_post/instagram_post listed but missing from canonical template, (7) non-self reviewer should confirm NO engagement metrics claimed, (8) bio disclosure = minimum bar not default — recommend in-body disclosure for future approvals. Draft assessed at 90% complete — the most thorough safety review of the matrix to date.