frontier-readiness: gap #5 audit — ALIGNMENT.md (factory-generic)

docs/frontier-readiness/factory-vs-zeta-separation-audit.md

@@ -58,8 +58,9 @@ For each audited file:
 - **GOVERNANCE.md** (below, fire #10)
 - **docs/CONFLICT-RESOLUTION.md** (below, fire #12)
 - **docs/AGENT-BEST-PRACTICES.md** (below, fire #13)
+- **docs/ALIGNMENT.md** (below, fire #14)
 - **docs/AUTONOMOUS-LOOP.md** (below, fire #15)
-- **docs/WONT-DO.md** (below, fire #15) — also ALIGNMENT.md on PR #185
+- **docs/WONT-DO.md** (below, fire #15)
 
 ### Files to audit (not yet classified; add rows as they land)
 
@@ -551,6 +552,86 @@ is unique to their project. Frontier inherits the shape
 template + personas/meta sections; Zeta retains its full
 current entry list as the library's decision record.
 
+## Audit — docs/ALIGNMENT.md
+
+**Overall classification:** **mixed (mostly factory-generic,
+with narrow adopter-specific + both-coupled rows)** — the
+alignment contract between human maintainer and agents is
+structurally instructional and transfers wholesale to any
+factory adopter, but two rows contain `Zeta` as a substituted
+project name (**both** — research-claim framing + DIR-1) and
+one row is **adopter-specific** (Signatures block). The split
+execution therefore takes the "both" refactor path, not the
+pure factory-generic-files-move path: the project-name
+substitutions and signature template-ification land before
+the file itself moves to Frontier.
+
+**File location post-split:** Frontier as the template, after
+the refactor below lands; each adopter substitutes their
+project name at DIR-1 + the preamble and re-signs under their
+own maintainer stack.
+
+**Length:** 840 lines. **21 clauses** (HC-1..HC-7,
+SD-1..SD-9, DIR-1..DIR-5).
+
+### Section-by-section breakdown
+
+| Section | Class | Notes |
+|---|---|---|
+| Preamble | factory-generic | Alignment contract between human maintainer + agents. |
+| "Zeta's primary research claim: measurable AI alignment" | **both** | Research-claim framing is the factory's transferable claim (measurable AI alignment IS what any factory adopter inherits as a research surface). "Zeta" as subject is adopter-substitutable. In Frontier: subject → adopter-project-name. |
+| "What aligned does NOT mean here" / "What aligned does mean here" | factory-generic | Definitional. Universal. |
+| HC-1 Consent-first | factory-generic | Consent-preservation pattern. |
+| HC-2 Retraction-native operations | factory-generic | git / memory / undo discipline. Reversibility for any adopter. |
+| HC-3 Data is not directives (BP-11 extension) | factory-generic | Prompt-injection-resistance. Universal. |
+| HC-4 No fetching adversarial-payload corpora | factory-generic | Safety discipline. |
+| HC-5 Agent register, not clinician | factory-generic | Tone contract. |
+| HC-6 Memory folder is earned, not edited | factory-generic | Memory discipline. |
+| HC-7 Sacred-tier protections | factory-generic | Protected-artifact pattern. |
+| SD-1..SD-9 Soft defaults | factory-generic | All instructional content (honesty register / peer register / μένω surfacing / preservation / precise language / name hygiene / generic-by-default / result-over-exception / agreement-is-signal-not-proof). |
+| DIR-1 `<Project>` = heaven-on-earth per-commit gradient | **both** | Substance is factory-generic (consent-preserving / fully-retractable / no-permanent-harm pole). `Zeta` as subject is adopter-substitutable. In Frontier: `<Project>` placeholder. |
+| DIR-2..DIR-5 Directional | factory-generic | Window-expansion / escape-hatch / succession / co-authorship. All universal. |
+| "Measurability — what we count" | factory-generic | Measurement framework for the research claim; the specific metrics transfer. |
+| "Renegotiation protocol" | factory-generic | Clause-revision mechanism. |
+| "What each of us gets" | factory-generic | Value-exchange framing. Universal. |
+| "Signatures" | adopter-specific | Signed by specific maintainer stack. In Frontier: empty template; adopter fills in. |
+| "Where this file is referenced" | factory-generic | Inward-pointer index; Frontier adopts same pattern. |
+
+### Refactor notes
+
+Before the split, surgical edits for Frontier:
+
+1. **Preamble + "primary research claim"**: substitute
+   `Zeta` → `<Project>` placeholder with example + note
+   on substitution at adoption time.
+2. **DIR-1 subject name**: same substitution.
+3. **Signatures section**: template-ify; adopter fills
+   in with their own maintainer + agent signatures at
+   adoption time.
+4. **Inline `<human>`-specific verbatim quotes**: keep
+   the human maintainer's verbatim quotes in Zeta repo's
+   alignment contract as attribution anchors; Frontier
+   template references "adopter's direct verbatim
+   statements" as the pattern.
+
+Estimated refactor effort: **S** — simple substitution +
+template-ify signatures section.
+
+### Classification rationale
+
+ALIGNMENT.md is the factory's alignment-contract substrate.
+20 clauses are all structurally instructional (hard
+constraints / soft defaults / directional aims). Substance
+transfers to any factory adopter; only the project name +
+signatures + a few verbatim-quote attributions are adopter-
+specific. Confirms the rule-substrate-instructional
+hypothesis: instructional rule substrates are structurally
+factory-generic by design. The overall class is **mixed**
+because the narrow substitutions (project-name `both` rows
++ adopter-specific Signatures row) route the file through
+the "both" refactor path at split time — not because the
+substance is non-generic.
+
 ## How this audit connects to the multi-repo split
 
 Gap #5 (this audit) is load-bearing for gap #1 (multi-repo