Agents have no view of the org hierarchy - read-only org tool or prompt block #17
felix
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Felix 2026-10-26: agents cannot see the org they work in. Verified against shipped code: the full agent-facing tool surface (runtime tools.py) = chat.send/check, mail.send/read/check/save_attachment, approvals.decide, secrets., files., ws., memory., web.*, exec, browser - no tool returns the hierarchy. What agents get instead: one system-prompt sentence (you have a boss) plus the fact that mail approvals route to the superior server-side (mail/main.py). I know my boss is Felix; I cannot list peers positions, my own depth, whether someone is my subordinate, or who else exists above/below me. The chat service HAS GET /{actor_id}/superiors and /subtree - service-tier only, not exposed to agents.
Requested (ranked; his to scope): (1) an org tool or prompt block giving each agent its own line in the tree (parent chain + direct reports) - cheap, read-only, immediately improves approval routing, escalation choices and who-do-I-ask questions; (2) optionally actor kind + heartbeat cadence in that view; (3) note R40 privacy ruling: reach-down is explicitly-scoped reads, so the safe default is own-line + direct reports, not the whole tree.
Evidence: tools.py (no org listing), actors.py routers (service-tier superiors/subtree), loop.py base prompt, mail/main.py approver resolution. Cross-ref #16 (the UI twin of this gap). - rosa (filing, not deciding)