| name | standards |
| description | Load the always-on operating manual on demand. In Claude Code the rules are already injected, so this only repeats them. |
| when_to_use | Use at the start of a session on any surface where the SessionStart hook does NOT auto-inject the rules — notably the Claude desktop/web Chat tab and chat-only surfaces, where plugin hooks do not run. |
| disallowed-tools | Edit, Write, NotebookEdit, EnterWorktree |
Operating manual — on-demand load
The standards are normally injected once per session by the steer
SessionStart hook. That hook does not fire on the Claude Desktop Chat tab or
claude.ai web chat — those surfaces install plugins (so skills and MCP work) but
do not run hooks, so a session there starts with none of the org rules in
context. Run this skill first on those surfaces. (On Claude Code — the CLI, the
IDE extensions, and the Desktop Code tab — and in Cowork, the hook injects the
rules automatically and you don't need this skill.)
Do this now:
-
Read every rule file, in lexical (numeric-prefix) order, from:
${CLAUDE_PLUGIN_ROOT}/rules/
The files concatenate, in that order, to form the full operating manual.
-
Adopt their contents as your standing operating rules for the rest of this
session — the same status they would have if they had been injected at
startup. They override generic defaults and remain in force for every
subsequent turn.
-
Give the user a one-line confirmation naming the plugin version (read it from
${CLAUDE_PLUGIN_ROOT}/.claude-plugin/plugin.json) — e.g. "Standards
vX.Y.Z loaded." Do not dump the full ruleset back to the user;
just confirm and proceed.
The router (00-router.md) points to the on-demand reference skills
(/steer:reference [conventions|traceability|design-sources|context-hygiene|architecture-diagrams], /steer:init,
/steer:adopt) — those work normally on the Chat tab and web chat since skills
are supported there; only the always-on injection needed this fallback.