From ea247bcf24d82630aef4ecac3f1e56281d37c045 Mon Sep 17 00:00:00 2001 From: Claude Date: Mon, 12 Oct 2026 06:08:42 +0000 Subject: [PATCH] docs: correct scheduled-routines.md's trigger tool-access claim (#654) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The doc claimed both cloud triggers run with tools limited to Bash, Read, Glob, Grep, Edit, Write, TodoWrite and no MCP tools, calling that the mechanical half of "provably excludes Layer 4". Querying the live trigger definitions (mcp__Claude_Code_Remote__get_trigger/list_triggers) shows neither carries an allowed_tools restriction: both have the full session tool surface plus nine MCP connectors with empty permitted_tools (unrestricted). There is no tool-level exclusion of Layer 4 — Bash alone can invoke it. The hermetic guarantee is enforced by the routine prompt's HARD RULES, not by tool access control. Co-Authored-By: Claude Sonnet 5 Claude-Session: https://claude.ai/code/session_01TgrUL5Je2LLNyxN61gE9qt --- doc/scheduled-routines.md | 19 +++++++++++++++---- 1 file changed, 15 insertions(+), 4 deletions(-) diff --git a/doc/scheduled-routines.md b/doc/scheduled-routines.md index a612314a7d..6b5440f554 100644 --- a/doc/scheduled-routines.md +++ b/doc/scheduled-routines.md @@ -48,10 +48,21 @@ connected for cloud sessions) is cleared, and both routines exist: | Weekly maintenance | `trig_01VgDQW3FWV5Ln4HUutsHBRt` | `0 6 * * 1` | | Monthly constitution audit | `trig_01RoxWz7WbEqtHxe4y62A99M` | `30 6 1-7 * 1` | -Both run `claude-sonnet-5` in the Default cloud environment against this repo, with -tools limited to `Bash, Read, Glob, Grep, Edit, Write, TodoWrite` — no browser or MCP -tools, which is the mechanical half of "provably excludes Layer 4". Manage them at -https://claude.ai/code/routines. +Both run `claude-sonnet-5` in the Default cloud environment against this repo. +**Corrected 2026-10-12 (issue #654):** there is no tool allowlist on either trigger. +Querying the trigger definitions directly (`mcp__Claude_Code_Remote__get_trigger` / +`list_triggers`) shows both carry the full session tool surface — `Bash`, `Agent`, +`Artifact`, `Workflow`, `ToolSearch`, `ScheduleWakeup`, `SendUserFile`, `Skill`, +`ListAgents`, the `github` MCP server, plus nine more MCP connectors (`Google_Calendar`, +`Slack`, `Gmail`, `Google_Drive`, `Figma`, `Vercel`, `Linear`, `Claude_Code_Remote`, +`Format`) each with an empty `permitted_tools` (i.e. unrestricted). There is **no +mechanical exclusion of Layer 4** — `Bash` alone is sufficient to invoke +`layer4cdp`/`dev-rig`/the sweeps. The "never run Layer-4, never message a live chat +host" guarantee is enforced entirely by this file's routine prompt (the HARD RULES +each fired session is instructed to follow), the same trust basis as every other +"never" in `AGENTS.md`, not a tool-level allowlist. Treat "hermetic" as a prompted +behavior, not a provable one, until/unless the trigger's `mcp_connections` are +actually pared down. Manage the triggers at https://claude.ai/code/routines. To (re)create one, use the `RemoteTrigger` tool with `action: "create"` and the prompt below. **The repository goes in `job_config.ccr.session_context.sources`**, as