The page described Inferencebot alone. It now documents the fleet: Inferencebot (steward), infrastructure-bot, socialbot and designbot, with a table of each agent's job, room, who may instruct it, and its access ceiling. Adds what they share (one Unix account each, bot accounts cannot use sudo, own gateway key and budget enforced at the gateway, no member data, room allowlist), the permission ladder (operator root > steward's bounded sudo to bot accounts > bots with no sudo), and the current instruction rules (mention required; room membership or named allowlist).
Also corrects the stale '#inferencebot-updates' claim (member-facing digests are posted by @inference-admin with an attribution line; the steward answers only in HQ) and the matching accounts section on Infrastructure, which said the admin account never speaks in a member room.
Verified against the live host: four Unix accounts and four gateway units; bot sudo refuses; LiteLLM teams agent: at $5/30d with the GreenPT/Tinfoil allowlist; env allowlists and rooms read from each agent's own config; MkDocs build renders the table as 5 columns x 4 rows.
The page described Inferencebot alone. It now documents the fleet: Inferencebot (steward), infrastructure-bot, socialbot and designbot, with a table of each agent's job, room, who may instruct it, and its access ceiling. Adds what they share (one Unix account each, bot accounts cannot use sudo, own gateway key and budget enforced at the gateway, no member data, room allowlist), the permission ladder (operator root > steward's bounded sudo to bot accounts > bots with no sudo), and the current instruction rules (mention required; room membership or named allowlist).
Also corrects the stale '#inferencebot-updates' claim (member-facing digests are posted by @inference-admin with an attribution line; the steward answers only in HQ) and the matching accounts section on Infrastructure, which said the admin account never speaks in a member room.
Verified against the live host: four Unix accounts and four gateway units; bot sudo refuses; LiteLLM teams agent:<bot> at $5/30d with the GreenPT/Tinfoil allowlist; env allowlists and rooms read from each agent's own config; MkDocs build renders the table as 5 columns x 4 rows.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
The page described Inferencebot alone. It now documents the fleet: Inferencebot (steward), infrastructure-bot, socialbot and designbot, with a table of each agent's job, room, who may instruct it, and its access ceiling. Adds what they share (one Unix account each, bot accounts cannot use sudo, own gateway key and budget enforced at the gateway, no member data, room allowlist), the permission ladder (operator root > steward's bounded sudo to bot accounts > bots with no sudo), and the current instruction rules (mention required; room membership or named allowlist).
Also corrects the stale '#inferencebot-updates' claim (member-facing digests are posted by @inference-admin with an attribution line; the steward answers only in HQ) and the matching accounts section on Infrastructure, which said the admin account never speaks in a member room.
Verified against the live host: four Unix accounts and four gateway units; bot sudo refuses; LiteLLM teams agent: at $5/30d with the GreenPT/Tinfoil allowlist; env allowlists and rooms read from each agent's own config; MkDocs build renders the table as 5 columns x 4 rows.
caa98e9402to34ade7a882