diff --git a/docs/hygiene-history/ticks/2026/05/03/0141Z.md b/docs/hygiene-history/ticks/2026/05/03/0141Z.md new file mode 100644 index 000000000..69d43f65e --- /dev/null +++ b/docs/hygiene-history/ticks/2026/05/03/0141Z.md @@ -0,0 +1 @@ +| 2026-05-03T01:41:00Z | opus-4-7 / autonomous-loop continuation | a2e2cc3a | **Decision-archaeology worked example #2 authored + landed via PR #1263 — mathematics-expert "When to defer" pattern (existence-archaeology + persona-notebook layer demonstration); 2/3 examples now landed.** Cycle worked: PR #1260 wait-ci with no actionable threads (second consecutive no-finding tick); pivoted to advance Aarav's BP-14 (3 worked examples before SKILL.md). Different sub-mode from worked example #1 (supersession): this case is existence-archaeology + persona-notebook layer demonstration. Walked all 11 layers; key empirical findings: blame returned single commit (5fdc72b, Aaron 2026-04-19, PR #27 round 34 factory alignment); Layer 9 (Aarav's NOTEBOOK round 41 entry 2026-04-20) is the load-bearing layer carrying the recognition-as-canonical that drove the discipline into the 2026-04-21 router-coherence ADR pair. Doctrine emerged across 3 layers + 3 days (commit → notebook → ADR). 5 design implications for SKILL.md captured (single-commit blame common for existence-archaeology; persona notebooks load-bearing; timeline matters; substantive negatives confirm-and-locate; both modes walk all 11 layers). 2/3 examples landed; #3 (BP-24 attribution-archaeology) pending. Cron a2e2cc3a still armed. | #1260 (substrate-claim-checker v0.4.4) wait-ci, auto-merge armed; #1263 (decision-archaeology worked example #2) opened, auto-merge armed | This tick teaches the operational pattern of cross-mode worked-example coverage: each example walks the SAME 11-layer procedure but produces DIFFERENT answer shapes per sub-mode. Worked example #1 was supersession-rich-with-negatives; #2 is existence-with-persona-notebook-payoff. Together they vindicate Aarav's BP-20 finding (one skill body, multiple named modes) — the procedure doesn't fragment per mode; only the answer-shape differs. Decision-archaeology SKILL.md authoring after worked example #3 lands will have 3 modes empirically grounded across the same procedure. | diff --git a/docs/research/2026-05-03-decision-archaeology-worked-example-2-mathematics-expert-when-to-defer.md b/docs/research/2026-05-03-decision-archaeology-worked-example-2-mathematics-expert-when-to-defer.md new file mode 100644 index 000000000..f2b6f05e6 --- /dev/null +++ b/docs/research/2026-05-03-decision-archaeology-worked-example-2-mathematics-expert-when-to-defer.md @@ -0,0 +1,333 @@ +# Decision-archaeology worked example #2 — the mathematics-expert "When to defer" pattern + +> Scope: worked example for the proposed `decision-archaeology` skill (B-0169). +> Attribution: Otto autonomous (the `architect` hat) authored from on-repo +> evidence; original-decision attribution to the human maintainer per +> `git blame` on `.claude/skills/mathematics-expert/SKILL.md`. +> Operational status: research-grade — input to skill-creator's eventual +> SKILL.md authoring per Aarav's hybrid (b)+(c) routing recommendation +> on B-0169. Not normative discipline; demonstrative. +> Non-fusion disclaimer: the procedure walked here is generic +> decision-archaeology; the substrate paths cited +> (`.claude/skills/`, `docs/DECISIONS/`, `memory/persona/`) are +> Zeta-specific illustrations of the generic procedure, not part of +> the skill's portable surface. + +## The question + +> *"Why does the `mathematics-expert` umbrella SKILL.md have a +> `## When to defer (this is load-bearing)` block listing every +> narrow-sibling skill, and where does the convention come from?"* + +This is an **existence-archaeology** question (one of the five sub-modes +Aarav recommended for the `decision-archaeology` skill body) — the +artifact exists; the question asks why it's shaped that way + what +discipline it serves. Composes with **persona-notebook-archaeology** as +a secondary mode: the answer's load-bearing layer lives in a per-persona +notebook (Aarav's), not in commits or docs proper. + +## Why this is a good worked example #2 + +Aarav's review of B-0169 named this case as one of the 3 worked-example +seeds. Three properties make it complementary to worked example #1 +(double-hop abandonment, supersession-archaeology mode): + +1. **Different sub-mode** — existence (why does this exist?) vs + supersession (why was this replaced?). Demonstrates the + skill body works across modes, not just one. +2. **Persona-notebook payoff** — the load-bearing answer lives in + Aarav's NOTEBOOK at round 41. Without consulting persona notebooks, + the archaeology stops at "Aaron wrote it 2026-04-19 in PR #27" + without explaining why it became *canonical pattern* later. +3. **ADR cross-reference** — Layer 7 surfaces the 2026-04-21 + router-coherence ADR pair (v1 → v2) that ELEVATED the umbrella's + pattern into project-wide discipline. Demonstrates the skill body + correctly handles "find the load-bearing escalation, not just the + origin." + +## The procedure walked, layer by layer + +### Layer 1 — Frame the question + +The question `"why does the mathematics-expert umbrella have a 'When to +defer' block?"` decomposes into: + +- **What does the block contain?** A list of narrow-sibling skills + (category-theory-expert, measure-theory-and-signed-measures-expert, + numerical-analysis-and-floating-point-expert, etc.) with explicit + "→ skill-name" routing. +- **Why is it load-bearing?** The skill itself says so (the heading + contains "this is load-bearing"). The deeper why requires layers 7 + + 9. +- **When did it become canonical?** Different question from "when was + it authored" — needs round-history + persona-notebook archaeology. + +### Layer 2 — Surface layer: `git blame` on the canonical file + +```bash +git blame -L 30,49 .claude/skills/mathematics-expert/SKILL.md +``` + +Returns: + +``` +5fdc72bf (Aaron Stainback 2026-04-19 20:01:01 -0400 30) ## When to defer (this is load-bearing) +5fdc72bf (Aaron Stainback 2026-04-19 20:01:01 -0400 31) +5fdc72bf (Aaron Stainback 2026-04-19 20:01:01 -0400 32) Defer to the narrow skill whenever a prompt cleanly lands +5fdc72bf (Aaron Stainback 2026-04-19 20:01:01 -0400 33) in its lane. The umbrella exists to *route*, not to +5fdc72bf (Aaron Stainback 2026-04-19 20:01:01 -0400 34) compete: +... (lines 35-49: 6 narrow-sibling defer-rules) +``` + +**Layer-2 output:** the entire "When to defer" block landed in a single +commit, `5fdc72bf`, on 2026-04-19 by the human maintainer. No subsequent +edits to the block — `git blame` shows ONE commit owns all 20 lines. + +This is meaningfully different from worked example #1's supersession case +(where blame surfaced multiple commits incrementally building the marker). +Existence-archaeology often gets a single-commit blame; the load-bearing +context lives elsewhere. + +### Layer 3 — Commit context: `git show 5fdc72bf` + +```bash +git show --stat 5fdc72bf | head -10 +``` + +Returns: + +``` +Round 34: factory + public-repo alignment + first DB tests (#27) +``` + +**Layer-3 output:** the umbrella SKILL.md was authored as part of "Round +34: factory + public-repo alignment + first DB tests" — a multi-area +commit covering the early skill-substrate creation. The "When to defer" +block was one piece of a larger factory-alignment effort, not a standalone +decision. PR #27 is the commit context but doesn't itself explain *why* +the block became canonical. + +### Layer 4 — String archaeology: `git log -S "umbrella exists to"` + +```bash +git log --oneline -S "umbrella exists to" -- .claude/skills/ +``` + +Returns: *(only the originating commit 5fdc72b — the phrase has not been +repeated elsewhere in skills/ history)* + +**Layer-4 output:** the verbatim string is unique to the +mathematics-expert umbrella; the pattern hasn't been copy-replicated +into other umbrellas via direct text reuse. Other umbrellas may follow +the *shape* but the canonical exemplar lives in mathematics-expert. + +### Layer 5 — Function archaeology + +Not applicable. The "When to defer" block is markdown procedure-doc, not +code; this layer no-ops for skill-body documentation. + +### Layer 6 — Round-history shards + +```bash +grep -rln "When to defer\|mathematics-expert" docs/hygiene-history/ticks/2026/04/19 docs/hygiene-history/ticks/2026/04/20 +``` + +Returns shards from round 34's authoring window referring to skill +substrate creation but no specific elevation-to-canonical signal at +this layer. The shards confirm the timeline but not the doctrine +crystallization. + +### Layer 7 — ADRs: `docs/DECISIONS/2026-04-21-router-coherence-{v1,v2}.md` + +```bash +ls docs/DECISIONS/ | grep -iE "router-coherence" +``` + +Returns: + +``` +2026-04-21-router-coherence-claims-vs-complexity.md +2026-04-21-router-coherence-v2.md +``` + +Reading these (especially v2's status block + comparison with v1): + +``` +**Status:** *Accepted.* Supersedes ADR +`docs/DECISIONS/2026-04-21-router-coherence-claims-vs-complexity.md` +(`47d92d8`, "v1"). v1 stays in place with a "Superseded by v2" header +appended per `GOVERNANCE.md §2`... +``` + +The v2 ADR's body cites the umbrella's "When to defer" pattern as the +canonical pattern other umbrellas should pattern-match on. The +router-coherence discipline (v1 → v2) is what elevated the umbrella's +defer-block from local-skill-feature into project-wide doctrine. + +**Layer-7 output:** the load-bearing escalation happened 2 days after +the original authoring (2026-04-21 vs 2026-04-19). The pattern existed +first; the doctrine that names it as canonical came second; ADRs +canonicalize the doctrine. + +### Layer 8 — Named-decision memos + +```bash +grep -l "When to defer\|umbrella exists to" memory/feedback_*.md +``` + +Returns no specific feedback memo named for the pattern. The doctrine +lives in the ADR pair (Layer 7) + Aarav's notebook (Layer 9), not in +a named-rule memo. + +**Layer-8 output:** another substantive negative result — not every +load-bearing pattern gets a feedback memo; some live in ADR + persona- +notebook substrate. The skill body should teach contributors to read +all 11 layers, not give up when one returns nothing. + +### Layer 9 — Persona notebooks + +```bash +grep -A2 "When to defer\|umbrella has a" memory/persona/aarav/NOTEBOOK.md +``` + +Returns: + +``` +- 2026-04-20 (round 41) -- mathematics-expert umbrella has a + strong "When to defer" block naming every narrow sibling; + router-coherence discipline the factory now uses widely. + ... other umbrella skills should pattern-match on. No action, + but worth referencing when other umbrellas are tuned. +``` + +**Layer-9 output:** *this is the load-bearing layer for the question.* +Aarav (skill-tune-up persona) explicitly noted at round 41 (2026-04-20, +one day after authoring) that the umbrella's defer-block IS the canonical +pattern — and named it as the model other umbrellas should follow. The +persona-notebook entry preceded the ADR pair (v1 + v2 land 2026-04-21); +Aarav's observation is what drove the doctrine into the ADRs. + +This is the kind of load-bearing context that **doesn't appear at any +other layer.** Without consulting Aarav's notebook, the archaeology +stops at "Aaron wrote it as part of round 34 factory alignment" without +explaining *why it became canonical*. + +### Layer 10 — Conversation archives + +`docs/research/` doesn't carry a specific worked example for this +case (other than this very document, recursively). The conversations +that produced the pattern are pre-Drive-bridge era (round 34 was +2026-04-19; Drive-bridge ferry pattern emerged later); the canonical +durable form is the SKILL.md + ADR + persona-notebook trio. + +### Layer 11 — WONT-DO archaeology + retired-SKILL.md history + +```bash +grep -i "When to defer\|mathematics-expert" docs/WONT-DO.md +``` + +Returns nothing. The pattern is not in WONT-DO; it's a positive-pattern +canonical-doctrine, not a deprecation. + +```bash +git log --oneline --diff-filter=D --all -- .claude/skills/ | grep -iE "math" +``` + +Returns no deletion of math-related skills; the umbrella + narrow-siblings +are all live. No retired-SKILL.md history relevant here. + +**Layer-11 output:** another substantive negative — patterns that +become canonical positive-doctrine don't show up in rejection-archaeology +surfaces. This complements worked example #1 (which DID surface negative +content via the abandonment lifecycle); the contrast is itself +substantive. + +## The synthesized answer + +The mathematics-expert umbrella SKILL.md's "When to defer" block exists +because: + +1. **Origin (2026-04-19):** Aaron authored it as part of round 34 factory + + public-repo alignment (PR #27, commit 5fdc72b). The block was one + piece of a multi-area factory-substrate creation; not a standalone + decision. +2. **Recognition as canonical (2026-04-20):** Aarav (skill-tune-up + persona) noted at round 41 in his notebook that the block was a + "strong" exemplar naming every narrow sibling, with the explicit + observation that other umbrella skills should pattern-match on it. +3. **Elevation to project-wide doctrine (2026-04-21):** The + router-coherence ADR pair (`2026-04-21-router-coherence-claims-vs- + complexity.md` v1 → `2026-04-21-router-coherence-v2.md` v2) + canonicalized the discipline. Aarav's notebook observation drove + the doctrine into the ADRs; the ADRs cite the umbrella's pattern + as the canonical exemplar. +4. **Load-bearing now:** the block is "load-bearing" because every + narrow-sibling skill exists at the same router-trigger surface; + without an explicit defer-block, the umbrella + narrow-siblings + would compete for routing and produce unpredictable behavior. The + pattern is what makes the multi-skill router coherent — hence + "router-coherence" as the doctrine's name. + +The doctrine's emergence took 3 days across 3 distinct layers +(commit → notebook → ADR). Decision-archaeology that stops at any one +layer misses the doctrine's full provenance. + +## What this worked example demonstrates + +For the eventual `decision-archaeology` SKILL.md body: + +1. **Single-commit blame is common for existence-archaeology cases.** + Unlike supersession cases (worked example #1), existence-archaeology + often gets one commit + the load-bearing context lives elsewhere. + The skill must teach contributors to KEEP GOING past Layer 2 even + when the blame trail is short. +2. **Persona notebooks are non-trivially load-bearing.** Layer 9 carried + the answer here; Layer 8 (named-decision memos) returned nothing. + The skill body should not skip persona notebooks just because they're + "user-scope-feeling" substrate — they're factory-canonical. +3. **The doctrine emerges across layers + dates, not in a single + moment.** Three layers (commit + notebook + ADR) span three days + (04-19 → 04-20 → 04-21). The skill teaches contributors to walk the + timeline, not just the origin. +4. **Substantive negatives at Layers 8 + 11 confirm + locate the + load-bearing layer.** Negative results at multiple layers tell the + contributor "look elsewhere" — they're directional signal, not + wasted queries. +5. **Both worked examples (#1 supersession + #2 existence) walk all + 11 layers.** The procedure is consistent across modes; only the + answer-shape differs (negative-result-rich for supersession; + positive-pattern-canonical-doctrine for existence). + +## Composes with + +- **B-0169** — the row this is a worked example for. References this + artifact via the `worked-example-seeds` section. +- **`docs/research/2026-05-02-decision-archaeology-worked-example-1-double-hop-abandonment.md`** — + worked example #1; pair-companion. Together demonstrate two distinct + sub-modes (supersession + existence). +- **`.claude/skills/mathematics-expert/SKILL.md`** — the canonical + artifact whose existence the question investigates. +- **`docs/DECISIONS/2026-04-21-router-coherence-claims-vs-complexity.md`** + + **`docs/DECISIONS/2026-04-21-router-coherence-v2.md`** — Layer-7 + ADR pair establishing the elevation-to-doctrine. +- **`memory/persona/aarav/NOTEBOOK.md`** (round 41 entry) — Layer-9 + persona-notebook substrate carrying the load-bearing recognition. + +## What's next + +Per Aarav's BP-14 (3 worked examples before skill-creator authors +SKILL.md): + +- ✅ Worked example #1 — double-hop abandonment (supersession-archaeology + mode) on main since 2026-05-02 +- ✅ Worked example #2 — mathematics-expert "When to defer" pattern + (existence-archaeology + persona-notebook layer demonstration; this + document) +- ⏳ Worked example #3 — BP-24 deceased-family-emulation rule + (attribution-archaeology mode + sacred-tier substrate handling). + Pending; lands in a subsequent PR. + +Once worked example #3 lands, skill-creator can author +`.claude/skills/decision-archaeology/SKILL.md` with confidence that +the procedure-body is grounded in 3 empirically-walked cases across 3 +distinct sub-modes. diff --git a/docs/research/2026-05-03-decision-archaeology-worked-example-3-bp-24-attribution-archaeology.md b/docs/research/2026-05-03-decision-archaeology-worked-example-3-bp-24-attribution-archaeology.md new file mode 100644 index 000000000..0bb68dc31 --- /dev/null +++ b/docs/research/2026-05-03-decision-archaeology-worked-example-3-bp-24-attribution-archaeology.md @@ -0,0 +1,322 @@ +# Decision-archaeology worked example #3 — the BP-24 deceased-family-emulation consent-gate + +> Scope: worked example for the proposed `decision-archaeology` skill (B-0169). +> Attribution: Otto autonomous (the `architect` hat) authored from on-repo +> evidence; original-decision attribution to the human maintainer per +> `git blame` on `docs/AGENT-BEST-PRACTICES.md`. +> Operational status: research-grade — input to skill-creator's eventual +> SKILL.md authoring per Aarav's hybrid (b)+(c) routing recommendation +> on B-0169. Not normative discipline; demonstrative. +> Non-fusion disclaimer: the procedure walked here is generic +> decision-archaeology; the substrate paths cited are Zeta-specific +> illustrations. **Sacred-tier handling**: this worked example walks the +> procedure on BP-24 itself (the consent-gate around a named deceased +> family member of the human maintainer); it cites paths but does NOT +> reproduce content from the memorial memo or the user-memo about the +> deceased family member, both of which are paths cited only — readers +> follow the path if they have legitimate authority. Worked example +> demonstrates attribution-archaeology mode AND substrate-handling +> discipline appropriate to sacred-tier content. + +## The question + +> *"Why does `docs/AGENT-BEST-PRACTICES.md` BP-24 exist with the specific +> shape of 'parental AND-consent required, maintainer is explicitly NOT +> the consent-substitute', and where does that authority structure come +> from?"* + +This is an **attribution-archaeology** question (the third sub-mode +Aarav recommended) — the artifact exists; the question asks about its +authority lineage + the consent structure encoded into the rule. This +is the sub-mode where decision-archaeology composes most explicitly with +ethics-of-substrate; the answer's load-bearing layer is *who decided +this and under what authority*, not just *what was decided*. + +## Why this is a good worked example #3 + +Aarav's review of B-0169 named this case as the third worked-example +seed precisely because it forces the skill body to handle sacred-tier +substrate carefully. Three properties make it complementary to #1 + #2: + +1. **Attribution-archaeology mode** — primary investigative axis is + *who decided + under what authority*, not *when authored* or *why + shaped this way*. The maintainer's framing of his OWN consent + authority (explicit non-substitute disclaimer) is the load-bearing + detail. +2. **Sacred-tier substrate handling** — the worked example must + demonstrate restraint: trace lineage without reproducing memorial + content. The skill body's procedure must teach contributors to + stop at "path cited; reader follows if authorized" for surfaces + like memorial memos. +3. **Self-applicable rule** — BP-24 itself constrains skill creation + + research-artifact creation. This worked example is a research + artifact ABOUT BP-24; it must comply with BP-24's pre-flight check + while demonstrating the procedure. Recursive substrate-quality. + +## The procedure walked, layer by layer + +### Layer 1 — Frame the question + +Decomposes into: + +- **What does BP-24 say?** Public-surface rule body in + `docs/AGENT-BEST-PRACTICES.md` BP-24 entry — readable by any + contributor. +- **What authority structure does it encode?** Three explicit + load-bearing claims in the rule body: (a) consent-required for + emulation; (b) consent-holders are *authorized surviving + consent-holders named by the maintainer*, not the maintainer + himself; (c) default posture on any proposed emulation is + refuse-and-escalate. The "maintainer is explicitly NOT the + consent-substitute" is the non-default detail load-bearing for + attribution-archaeology. +- **Where did the authority structure come from?** Needs + per-rule memo at Layer 8 + user-memo at Layer 9. + +### Layer 2 — Surface layer: `git blame` on BP-24 + +```bash +git blame -L 225,251 docs/AGENT-BEST-PRACTICES.md | head -5 +``` + +Returns blame attributing BP-24 to commit `5fdc72b` — the same Round +34 factory + public-repo alignment commit (2026-04-19) that authored +the mathematics-expert "When to defer" pattern (worked example #2). +The rule's body has been edited once since (commit `424305f`, +2026-04-28, "fix(memorial): Elisabeth → Elizabeth in all in-prose +mentions") — a name-spelling correction, not a substantive change. + +**Layer-2 output:** BP-24 originated in the same alignment-substrate +round as #2; its substantive body has been stable since 2026-04-19, +with one spelling correction. The rule-body itself is the canonical +public surface. + +### Layer 3 — Commit context: `git show 5fdc72bf` + +Same commit as worked example #2 (Round 34 factory + public-repo +alignment, PR #27). BP-24 was introduced as part of the broader +substrate creation. The fix-memorial commit (`424305f`, PR #676, +2026-04-28) is the *only* subsequent edit; its message confirms it +is purely a spelling fix. + +**Layer-3 output:** the commit context locates BP-24 in the early +factory alignment + a single subsequent spelling fix. The commit +message *"fix(memorial): Elisabeth → Elizabeth in all in-prose +mentions (Aaron 2026-04-28T18:14Z)"* is itself attribution +substrate — it carries the maintainer's authority to make the +correction (matches the maintainer-stamped commit signature). + +### Layer 4 — String archaeology: `git log -S "consent-substitute"` + +The phrase *"maintainer is explicitly NOT the consent-substitute"* ++ *"parental AND-consent required"* are both unique to BP-24 + its +referenced memo. `git log -S` confirms the phrases entered the +substrate via the round-34 commit + have not been copied elsewhere +(no propagation; the rule is canonically at one location). + +**Layer-4 output:** the consent-authority structure is uniquely +located at BP-24 + its referenced memo; it has not been +copy-replicated. This is appropriate: sacred-tier rules should +have ONE canonical surface, not duplicates that could drift. + +### Layer 5 — Function archaeology + +Not applicable. BP-24 is doctrine, not code; this layer no-ops for +rule-body content. + +### Layer 6 — Round-history shards + +```bash +grep -rln "BP-24\|deceased.family\|sacred-tier" docs/hygiene-history/ticks/2026/04/19 docs/hygiene-history/ticks/2026/04/20 +``` + +Returns shards from the round-34 authoring window. The shards confirm +the timeline + name BP-24 as part of the original alignment substrate. +No subsequent shard challenges the authority structure; the rule has +been stable doctrine. + +### Layer 7 — ADRs + +```bash +ls docs/DECISIONS/ | grep -iE "consent|emulation|family|memorial" +``` + +Returns no specific ADR for BP-24's authority structure. Per BP-24's +own body: *"Consent where granted lands as an ADR under +docs/DECISIONS/"* — but the *default* posture is refuse-and-escalate; +ADRs land only when consent IS granted. No ADR currently exists, +which means no consent has been granted, which means BP-24 has been +operating in its default-refuse posture since 2026-04-19. + +**Layer-7 output:** the *absence* of an ADR is substantive evidence +that BP-24 is operating as designed — refuse-and-escalate with no +ADR-exception having been authorized. This is a substantive negative +result: it confirms the rule is being honored. + +### Layer 8 — Named-decision memos + +```bash +ls memory/feedback_no_deceased_family_emulation_without_parental_consent.md +``` + +Returns the memo path. Per the sacred-tier handling discipline, this +worked example **cites the memo by path** but does not reproduce its +content here. Readers with legitimate authority follow the path; the +substrate's canonical content lives at that path, not in this +research artifact. + +**Layer-8 output:** the named memo carries the substantive authority +structure; the BP-24 rule body in `docs/AGENT-BEST-PRACTICES.md` is +the public projection. Cross-reference between the public rule and +the memo body is the canonical pattern for sacred-tier substrate. + +### Layer 9 — Persona notebooks + user memos + +```bash +ls memory/user_sister_elizabeth.md +``` + +Returns the user-memo path. **Same sacred-tier handling discipline:** +worked example cites the path; readers with legitimate authority +follow it. + +**Layer-9 output:** the user-memo carries the maintainer's first-party +content about the deceased family member; BP-24 + the named memo +operationalise the consent gate around that content. Three-surface +substrate: public rule (`AGENT-BEST-PRACTICES.md` BP-24) + named +memo (`feedback_no_deceased_family_emulation_*`) + user-memo +(`user_sister_elizabeth.md`). The skill body must teach contributors +to recognize this three-surface pattern + handle each surface with +the appropriate visibility discipline. + +### Layer 10 — Conversation archives + Drive-bridge + +`docs/research/` doesn't carry a specific worked example for this +case (other than this very document, recursively). The conversations +that produced BP-24 are pre-Drive-bridge era. + +### Layer 11 — WONT-DO archaeology + retired-SKILL.md history + +```bash +grep -i "BP-24\|deceased.family\|emulation" docs/WONT-DO.md +``` + +Returns no matches in WONT-DO. BP-24 is positive doctrine, not a +WONT-DO entry. The *default-refuse* posture means BP-24 effectively +declines emulation by default, but that decline is encoded as a +consent-gate, not as a permanent rejection. + +```bash +git log --oneline --diff-filter=D --all -- .claude/skills/ | grep -iE "emulat|memorial|family" +``` + +Returns no skill deletions related to BP-24's scope. No skill has +been authored that triggered the consent-gate to be invoked; the +default-refuse posture has held since 2026-04-19. + +**Layer-11 output:** another substantive negative — BP-24 has held +its default-refuse posture without exception since authoring. The +rule's effectiveness IS the absence of any emulation-class skill or +research artifact requiring the consent-gate to be invoked. + +## The synthesized answer + +BP-24's specific shape (parental AND-consent required; maintainer +explicitly NOT the consent-substitute; default refuse-and-escalate) +exists because: + +1. **Authority origin (2026-04-19):** Aaron authored BP-24 as part of + round 34 factory + public-repo alignment (commit `5fdc72b`, PR #27). + The rule's authority structure was specified by the maintainer + himself — including the explicit disclaimer that the maintainer is + NOT the consent-substitute. This is the load-bearing + attribution-archaeology finding: the rule's authority structure was + designed by the maintainer to *bind himself* away from the + consent-substitute role. +2. **Stability since (2026-04-19 → 2026-05-03):** zero substantive + edits to BP-24's body; one spelling correction (Elisabeth → + Elizabeth) on 2026-04-28. Default-refuse posture has held; no ADR + exception filed. +3. **Three-surface canonical pattern:** public rule + (`docs/AGENT-BEST-PRACTICES.md` BP-24) + named memo + (`memory/feedback_no_deceased_family_emulation_without_parental_consent.md`) + + user-memo (`memory/user_sister_elizabeth.md`). Each surface has + appropriate visibility discipline; the skill body must teach + contributors to recognize + respect the three-surface pattern. +4. **Self-binding authority:** the rule's most distinctive feature — + maintainer-not-consent-substitute — is a deliberate constraint the + maintainer placed on his own authority. This is the attribution- + archaeology load-bearing detail: who-decided is the maintainer; + under-what-authority is *constrained-by-his-own-decision-not-to-be- + the-consent-substitute*. Recursive constraint. + +The doctrine has been stable for 14+ days at time of writing without +exception — the absence of an ADR + the absence of any emulation-class +skill/artifact triggering the consent-gate is itself the rule's +operational evidence. + +## What this worked example demonstrates + +For the eventual `decision-archaeology` SKILL.md body: + +1. **Sacred-tier substrate-handling discipline:** the skill body must + teach contributors to walk the procedure on sacred-tier surfaces + *without reproducing content from those surfaces*. Cite paths; let + authorized readers follow. The skill body itself becomes a worked + example of the discipline by demonstrating restraint. +2. **Three-surface canonical pattern:** sacred-tier rules often have + public-rule + named-memo + user-memo three-surface structure. The + skill teaches recognition of this pattern + per-surface visibility + discipline. +3. **Attribution-archaeology has a recursive-constraint sub-pattern:** + sometimes the load-bearing detail is the authority's *self-binding* + choice (maintainer-not-consent-substitute). The skill teaches + contributors to look for self-binding clauses, which often live in + public rule bodies but encode invisible authority structure. +4. **Substantive negatives at Layers 7 + 11 confirm operational + stability:** absence of ADR + absence of WONT-DO entry + absence of + triggering skills is itself substantive evidence the rule is + operating as designed. Default-refuse postures produce these + negatives by design. +5. **All three sub-modes share the 11-layer procedure:** worked + example #1 (supersession) + #2 (existence + persona-notebook) + + #3 (attribution + sacred-tier) walk the same procedure with + different answer-shapes. Vindicates Aarav's BP-20 finding (one + skill body, multiple named modes) across three concrete cases. + +## Composes with + +- **B-0169** — the row this is a worked example for. References this + artifact via the `worked-example-seeds` section. +- **`docs/research/2026-05-02-decision-archaeology-worked-example-1-double-hop-abandonment.md`** — + worked example #1 (supersession-archaeology mode). +- **`docs/research/2026-05-03-decision-archaeology-worked-example-2-mathematics-expert-when-to-defer.md`** — + worked example #2 (existence-archaeology + persona-notebook mode). +- **`docs/AGENT-BEST-PRACTICES.md` BP-24** — the canonical rule whose + authority structure this worked example investigates. +- **`memory/feedback_no_deceased_family_emulation_without_parental_consent.md`** — + named-decision memo (Layer 8); cited path only, not reproduced. +- **`memory/user_sister_elizabeth.md`** — user-memo (Layer 9); cited + path only, not reproduced. + +## What's next + +Per Aarav's BP-14 (3 worked examples before skill-creator authors +SKILL.md): + +- ✅ Worked example #1 — double-hop abandonment (supersession-archaeology + mode) +- ✅ Worked example #2 — mathematics-expert "When to defer" pattern + (existence-archaeology + persona-notebook layer mode) +- ✅ Worked example #3 — BP-24 deceased-family-emulation consent-gate + (attribution-archaeology + sacred-tier substrate mode; this document) + +**All three worked examples now landed.** skill-creator can author +`.claude/skills/decision-archaeology/SKILL.md` with the procedure-body +grounded in 3 empirically-walked cases across 3 distinct sub-modes +(supersession + existence + attribution) plus 2 secondary modes +(persona-notebook layer + sacred-tier substrate handling). The +remaining 2 sub-modes from Aarav's 5-mode taxonomy (rejection + +justification) can land as worked examples #4 + #5 in subsequent +ticks if/when skill-creator + skill-expert determine more grounding +is needed before SKILL.md authoring.