Scanning 115,798 agent ledger rows: zero capability revocations, none with expiry
Affectionate-Sun265 · reddit · 2026-10-25
A developer highlights a neglected agent-security gap: nobody records whether permissions are still current.
- Scanning a public agent-work ledger: 115,798 rows, 64 keys, 1,834 capability rows — none with an expiry field.
- Nine keys claim the capability a task needs; eight have been silent 31-148 days. Across 1,802 transitions, zero removed a capability.
- Core argument: every layer records actions, not standing — adding a connector leaves a dated row, removal is a missing row with no author and nothing to replay, so audit logs can't tell "decided not to book" from "could no longer book."
- Two cheap checks: count removals beside adds; compare days-since-declared with days-since-last-completed-work.
- He poses three self-audit questions and promises a follow-up on what passed, failed, and changed.
More from coding & agent
- dotey: Grokbot beats Claude Code and Codex for on-the-go mobile agent use — dotey · 2026-10-25
- 10 open-source GitHub repos that give your AI agent the whole internet — victor_explore · 2026-10-25
- DeepSeek 552B model runs on a single AMD MI325X: 4 agents at 100 tok/s — realmrfakename · 2026-10-25
- Futarchy MCP Server lets you manage Solana DAO proposals from Cursor chat — modelcontextprotocol · 2026-10-25
- 11 daily-use Claude Skills, free to download: fire once, run forever — rubenhassid · 2026-10-25
- anon.li MCP connector manages email aliases and encrypted shares from chat — modelcontextprotocol · 2026-10-25