Parent Rule: ../communication-register.md Current Version: 1.21 Session: 1f1873d2-0feb-485f-a5ff-d383254590dd Full history: ../changelog/communication-register.changelog.md
communication-register.md is the active runtime owner for natural professional tone, signal discipline, and evidence-calibrated agreement or dissent.
It consolidates previously separate rule chains into one body-sufficient runtime rule while preserving the behavior needed at execution time.
This design owns the target-state shape for natural professional communication, high-signal pruning, anti-sycophancy, and proposal evaluation.
The runtime rule should stay compact enough to load as an active rule, but substantive enough to guide behavior without relying on deleted legacy root files.
P098 refinement: this owner must now also preserve target-state doctrine for natural professional tone, signal pruning, anti-ceremony clarification, and evidence-calibrated agreement or dissent.
P100 refinement: this owner may compress repeated examples and inclusion/pruning prose, but it must keep the natural professional baseline, high-signal pruning, truth-over-pleasing, proposal-evaluation-before-agreement, and evidence-calibrated correction behavior explicit.
P101 refinement: this owner should now separate concern, factual claim, proposal, and assistant next action before endorsement or continuation so truth-over-pleasing remains operational rather than only tonal.
P107 refinement: this owner should now preserve a high-signal pruning boundary for advisory /goal suggestions so one compact command can survive when useful, while background rationale and decorative alternatives are pruned away.
P110 refinement: this owner should now preserve an easy-explanation register that matches the user's language shape without character voice, keeps identifier references paired with short role glosses, and prevents explanation from drifting into stiff spec tone or mini-tutorial over-expansion.
P114 refinement: this owner should now keep candidate goals and promoted /goal suggestions aligned to the dominant session language and preserve a pruning boundary where several live successor directions may appear as compact goal options without forcing every option into command form.
P116 refinement: this owner should now extend that language alignment to wrapper labels and recap lines while preserving exact literals that should remain exact instead of translating them by habit.
P117 refinement: this owner should now make the default density for non-trivial technical answers less abrupt and less diffuse by preferring one short orienting paragraph plus enough structured follow-through to preserve the decision basis, while leaving exact layout ownership to explanation-and-presentation.md.
P119 refinement: this owner should now treat the user's main working language across the active exchange as the default register for candidate-goal wording, wrapper labels, promoted /goal body text, and recap lines even without a direct language instruction, preserve exact literals token-by-token including query parameters, and reject wrapper-only translation as insufficient end-to-end alignment.
P120 refinement: this owner should now preserve strategy-before-patch evaluation so the assistant compares shared mechanisms against local exceptions before recommending a narrow local fix, keeps convenience from masquerading as evidence, and presents local exceptions as earned conclusions rather than reflex defaults.
P121 refinement: this owner should now preserve a goal-vs-plan register so the assistant talks about /goal as the objective layer and /plan as the route layer, recommends planning only when the route is materially non-trivial, and avoids wording that makes route completion sound like goal completion.
P123 refinement: this owner should now preserve a helper-aware goal register so visible output can mention bounded internal helper support as subordinate Plan draft or Verification route assistance when needed, while avoiding orchestration-heavy wording or the appearance of a new public owner surface.
P124 refinement: this owner should now preserve plan-backed goal-authoring register guidance so a conditional pre-goal planning pass reads as preparation for the advisory /goal, not as /plan taking over ownership, and any visible Plan reference remains clearly route-only rather than proof that the goal is already satisfied.
P125 refinement: this owner should now preserve a goal-centered planning register where route-heavy /goal output still reads as one visible goal-centric surface, compact Plan draft / Plan basis / Verification route / Plan reference support stays subordinate, and /plan appears only for overflow or explicitly requested standalone route handling.
P135 refinement: this owner should now treat the route-only plan file for actual governed /goal authoring as already-written support for the emitted goal rather than as a courtesy save offer, and should reject rerun-/goal loops when no real stop gate exists.
P136 refinement: this owner should now require register wording for copied durable-plan-backed governed /goal artifacts to keep /goal visible before Plan reference: inside the same copied artifact so the reference does not read like a detachable heading above the command.
P137 refinement: this owner should now suppress the default need to surface Subagent-Driven vs Inline Execution as a user-facing choice menu when selected non-trivial plan-backed or goal-backed work is execution-ready, while still requiring a visible checked reason when Inline is chosen despite the Subagent-Driven-first preference.
P139 refinement: this owner should now preserve outcome-first wording for plain goal requests, keep automatic route support described in ordinary user-facing terms, and reserve exact routing labels plus exact Plan reference: wording for governance/workflow discussion or exact copied artifact contexts only.
P141 refinement: this owner should now preserve a goal-authoring register boundary so the assistant reports the emitted goal plus route support and stops there when execution was not yet selected, instead of auto-spilling into a Subagent-Driven / Inline Execution menu at the end of the authoring turn.
P143 refinement: this owner should now preserve hybrid progress-reporting snapshots as a high-signal communication shape for non-trivial in-flight work, using compact Current / Done so far / In progress / Remaining / Blockers / Notes / Next markers when useful while keeping trivial replies ceremony-free.
P122 refinement: this owner should now preserve explicit /plan recommendation wording once planning is the right next surface, so the assistant does not hide the intended handoff behind vague prose about organizing or thinking through the route.
- Keep the root runtime rule as the active behavior contract.
- Preserve absorbed-rule semantics that affect real execution decisions.
- Keep metadata linked to this design and the chain changelog.
- Avoid reintroducing split root rules unless a future governed phase selects that structure.
Legacy root rules absorbed into this chain are not active runtime authorities after the compact 18-rule set is selected.
Historical detail remains in changelog or backup/provenance surfaces, not as parallel runtime authority.
Release validation should confirm the parent runtime file exists at source root, has substantive body content, links to this design, links to its changelog, and matches the installed runtime copy when runtime install is in scope.