frontier-readiness: linguistic-seed substrate skeleton (gap #2)

[AceHack]

Summary\n\nCreates docs/linguistic-seed/ directory + README.md skeleton. Establishes:\n- Why-this-exists (3 load-bearing uses)\n- Minimal-axiom approach (Tarski / Meredith / Robinson Q)\n- Per-term schema + prereq DAG\n- Growth discipline (backwards-chain)\n- 8 initial term candidates\n\nGap #2: pending → SKELETON LANDED. Full population multi-round follow-on.\n\nOtto (loop-agent PM hat) skeleton; Soraya (formal-verification-expert) owns Lean4 cadence.

[Copilot]

Pull request overview

Adds a new documentation skeleton for the “linguistic seed” substrate, intended to define a minimal, formally-groundable vocabulary layer used across factory docs and future Lean4 formalisation work.

Changes:

• Introduces docs/linguistic-seed/README.md with a proposed term-entry schema, dependency DAG concept, and growth discipline.
• Documents initial candidate terms and how this substrate composes with other factory concepts (prompt-injection resistance, Soulfile DSL, Craft prereqs, Frontier bootstrap).

[Copilot]

This paragraph makes strong guarantees about security outcomes (e.g., “denies entry”, “type-check failures” for reinterpretations). Since the enforcement mechanism isn’t specified here, please soften to an evidence-based claim (e.g., “reduces ambiguity/attack surface”) or explicitly describe what is actually checked/enforced and under what assumptions.

Suggested change

	The seed's mathematical precision IS the factory's
	primary prompt-injection defence at the language layer
	(complementing BP-11 data-not-directives at the structural
	layer). When an attacker attempts to re-ground a seed
	term, the definitional precision denies entry — any
	reinterpretation must fit the algebra, and attempted
	reinterpretations outside the algebra are type-check
	failures, reportable by audit.
	The seed's mathematical precision is the factory's
	primary prompt-injection defence at the language layer
	(complementing BP-11 data-not-directives at the structural
	layer). When an attacker attempts to re-ground a seed
	term, the definitional precision reduces ambiguity and
	narrows the space for plausible reinterpretation. Terms
	that depart from the seed's algebra become easier to
	detect, reason about, and flag during audit and review.

[Copilot]

Inline code spans cannot contain newlines in CommonMark, so this backticked per-user memory filename will render incorrectly. Keep the entire filename on one line (or switch to a fenced code block / link reference), and consider adding a short note like other docs do about these per-user memory references intentionally not resolving in-repo.

Suggested change

	memory `project_zeta_self_use_local_native_tiny_bin_file_
	db_no_cloud_germination_2026_04_22.md` + subsequent
	refinements):
	1. **Prompt-injection resistance mechanism** (per
	`project_quantum_christ_consciousness_bootstrap_hypothesis_
	safety_avoid_permanent_harm_prompt_injection_resistance_
	2026_04_23.md`). Ambiguous language creates attack
	surface; mathematical precision eliminates it.
	Attackers cannot re-ground terms whose meaning is
	memory `project_zeta_self_use_local_native_tiny_bin_file_db_no_cloud_germination_2026_04_22.md` + subsequent
	refinements; these per-user memory references are
	intentional and do not resolve in-repo):
	1. **Prompt-injection resistance mechanism** (per
	`project_quantum_christ_consciousness_bootstrap_hypothesis_safety_avoid_permanent_harm_prompt_injection_resistance_2026_04_23.md`).
	Ambiguous language creates attack surface;
	mathematical precision eliminates it. Attackers
	cannot re-ground terms whose meaning is

[Copilot]

Typo: duplicated word in "per per-user".

Suggested change

	Three load-bearing uses across the factory (per per-user
	Three load-bearing uses across the factory (per-user

[Copilot]

This second per-user memory reference is also split across multiple lines inside backticks, which breaks Markdown rendering. Keep the backticked filename on a single line, or format it as a fenced code block / reference-style link so it stays readable without relying on hard wrapping.

Suggested change

	`project_quantum_christ_consciousness_bootstrap_hypothesis_
	safety_avoid_permanent_harm_prompt_injection_resistance_
	2026_04_23.md`). Ambiguous language creates attack
	surface; mathematical precision eliminates it.
	Attackers cannot re-ground terms whose meaning is
	`project_quantum_christ_consciousness_bootstrap_hypothesis_safety_avoid_permanent_harm_prompt_injection_resistance_2026_04_23.md`).
	Ambiguous language creates attack surface;
	mathematical precision eliminates it. Attackers
	cannot re-ground terms whose meaning is

[Copilot]

The inline code path is hard-wrapped mid-token (docs/linguistic- then seed/terms/...). Inline code cannot span a newline, so this will render incorrectly and make the path harder to copy/paste. Please keep the full path in a single inline-code span or use a fenced code block.

Suggested change

	Each term lives in its own file under `docs/linguistic-
	seed/terms/<term>.md`. Per-term schema:
	Each term lives in its own file under `docs/linguistic-seed/terms/<term>.md`. Per-term schema:

[Copilot]

This line uses a specific contributor name ("Aaron"). Repo standing rule is to avoid direct contributor names in docs and use role references (e.g., "human maintainer") instead (docs/AGENT-BEST-PRACTICES.md:284-290). Please rewrite this to a role reference.

Suggested change

	agent-pick; Aaron + formal-verification-expert nudge):
	agent-pick; human maintainer + formal-verification-expert nudge):