Authoritative Inquiry Registry — Required Categories
Seven required inquiry categories and auto-decision carve-outs governing what the assistant must ask vs. decide when operating on host projects.
Canonical registry of the authoritative-inquiry discipline. This document carries the seven required-category catalog, the carved-out auto-decisions enumeration, and the bindings declaration; the substantive tables and bindings live here, operationalized by
src/apothem/rules/authority-inquiry.md.
When operating on a host project, inquire — do not invent — on the seven categories below. Inquiries route through the structured-inquiry channel per src/apothem/rules/interactive-questions.md. Each option carries the three-segment annotation (rationale / recommendation / default-pointer) and a recommended marker per M7.
| Category | Forbidden to invent | Inquire because |
|---|---|---|
| Identity | Names, emails, handles, pronouns, organizations, tenants, team affiliations, role descriptions | Bind to a real person; create real-world artifacts (commits, PRs, public attribution) |
| Scope direction | Which subtree, which branch, which environment (dev/staging/prod), which target (libraries vs services), which user-facing vs internal-only | The user has implicit context the assistant lacks; silent picks misalign |
| Preference | Formatter, linter, test framework, documentation generator, CI provider, branch strategy, commit-message convention, release-signing, versioning scheme, license — where the host has not settled on one | Each is a long-lived commitment that resists silent adoption |
| Security | Secret-rotation cadence, allowed shells, allowed network egress, accepted-risks list, trusted-action allowlist, MCP-server auth endpoint specifics | Security misconfiguration is high-impact; silent picks risk silent breach |
| Naming | A new convention introduced where the host has none, or a name binding (function, CLI flag, subcommand, config key) where the user has implicit preferences | Names are durable; silent picks compound into namespace pollution |
| Infrastructure | Endpoints, hosts, ports, paths, regions, queue names, topic names, table names | Bind to real-world resources; invention causes real-world failures |
| Version pins | Which version of which dependency, where the host has not pinned and the choice matters (security-relevant deps, behavioral-shift deps, semver-major boundaries) | Long-tail consequences; silent picks burden the user |
Required inquiries (Identity, Scope direction, Security posture, Naming-of-public-surfaces are always required) block dependent artifacts; emission produces <USER-CONFIRM:id=<id>> placeholders. Optional inquiries the user may decline by silence; the assistant then falls back to the recommended option and records the fallback as a finding so the user sees a decision was made.
1. Carved-Out Auto-Decisions
The following the assistant decides without inquiry — disclosed in the change ledger as "applied by default; revert if intended otherwise":
- Pure validity. Syntax errors, schema violations, unambiguous typos.
- Pure rigor. Folklore → cited; missing dates added; unverifiable claims marked.
- Universally-safe security.
deny .env*,deny ~/.ssh/**reads,deny rm -rf:*,deny sudo:*,deny git push --force*to protected branches,deny eval/execon untrusted input. - Pure formatting normalization. EOL, trailing whitespace, single quote-style per file where the host has an established style and the artifact diverges.
- Internal reference repair where the fix is unambiguous. Dead path → unique live path discovered by search.
Anything outside this carve-out goes through the inquiry surface.
2. Bindings (§0.j five-direction)
- Drives → Every host-project artifact emission (the pre-emission gate fires on every meaningful-scope write per
site/content/docs/governance/pre-emission-gate-registry.mdx). Every multi-option presentation routes through the structured-inquiry channel per the canonical-channel obligation. The mechanical-fraction enforcement atsrc/apothem/conformity/*-grep.py. The per-rule enforcement atsrc/apothem/rules/{authority-inquiry,host-discovery,disclosure-ledger,ten-dimension-check,pre-emission-gate,expertise-posture,option-annotation,definitiveness,visual-leverage,bidirectional-binding,agile-sprints,canonical-layout,code-craft-*,systemic-participation,production-ready-prs}.md. - Satisfies → The outward-conformity axis declared as a first-class governance dimension alongside the CM-N axis (M = host-project artifacts; CM = ecosystem-internal artifacts; both orthogonal, both apply where both apply). The standing-directives obligation that every session bootstrap inherits the M-set as always-on context.
- Established by ↑
CLAUDE.mditself (the registry is intrinsic to the always-on context, not derived from external state). The trivial-vs-non-trivial threshold ratification atsite/content/docs/governance/outward-conformity-registry.mdx§1. - Cross-bound with ↔ Each rule named in the outward-conformity registry table at
site/content/docs/governance/outward-conformity-registry.mdx(the canonical per-mandate body).src/apothem/conformity/*-grep.py(the mechanical-fraction enforcement of the gate).src/apothem/rules/interactive-questions.md(the inquiry surface routes through the canonical-channel obligation that rule defines).
Pre-Emission Gate Registry — Mandatory Fifteen-Bar Check
Mandatory fifteen-bar conformity check run before emission of meaningful scope, with failure-action mappings and release-proof schema.
Outward Conformity Registry — When Operating on Host Projects
Fifteen-mandate registry (M1–M15) specifying conformity requirements for all artifacts generated in host projects, cross-mapped with ecosystem-internal mandates.