All 176 files in the four accumulating directories now use a zero-padded 4-digit counter prefix that reflects creation order (`NNNN-slug.md`). The counter is assigned per directory in strict git-log creation order; ties broken alphabetically by original name. The old `YYYY-MM-DD-` prefix on docs/specs/ and docs/plans/ files is dropped — the date is recoverable from git log and the counter carries the ordering. A file's counter is stable for the life of the file: never reassigned, never reused, never compacted. Deleted files retire their counter; subsequent files do not fill the gap. This is the property that lets cross-references stay literal — refs use the full filename including the counter (`design/contracts/0007-honesty-rule.md`) so they grep cleanly and resolve directly without a glob step. 313 cross-references updated across .md/.rs/.toml/.c/.json files (test pins, include_str! paths, design-INDEX entries, baseline notes, runtime C comments, inter-contract markdown links incl. bare basename and `../models/foo.md` forms). CLAUDE.md gets a new "File-naming convention" section spelling out the rule and rationale. skills/brainstorm/SKILL.md and skills/planner/SKILL.md updated so new spec/plan creation produces counter-prefixed names from the start. The full test suite (cargo test --workspace) passes.
10 KiB
Embedding ABI — M2.tidy — Implementation Plan
Parent spec: none — audit-driven docs-honesty tidy. Source:
docs/journals/2026-05-18-audit-embedding-abi-m2.md(committedad88dec), the M2 milestone-close audit's one actionable drift item ([medium]DESIGN.md:2266 stale(M1)header + coupled[low]:2354 xref), plus a Boss-added 5th live-coupling site the recon surfaced (form_a.md:89).For agentic workers: REQUIRED SUB-SKILL: use
skills/implementto run this plan. Steps use- [ ]checkboxes for tracking.
Goal: Rename the DESIGN.md section ## Embedding ABI (M1) →
## Embedding ABI (current-state-mirror — its body now describes the
post-M2 ctx ABI present-tense) and move every live coupled
reference in lockstep, completely, so no pointer dangles at a vanished
section title.
Architecture: Pure docs/test-string tidy. Five live-coupling edit
sites across three files (DESIGN.md header + DESIGN.md schema-block
xref comment + docs_honesty_pin.rs test comment + that test's
assert-message + form_a.md forward-xref). Zero language / checker /
codegen / schema change. The docs_honesty_pin.rs:135 asserted
substring (the DESIGN.md body sentence "Export parameters are written
bare…", DESIGN.md:2297–2299) is rename-immune and stays
byte-verbatim untouched — recon proved it lives 31 lines below the
header, is section-name-free, and is matched file-wide via
d.contains, so a header rename cannot affect it.
Tech Stack: docs/DESIGN.md, crates/ailang-core/tests/docs_honesty_pin.rs,
crates/ailang-core/specs/form_a.md (the latter is compiled in via
ailang_core::FORM_A_SPEC include_str!). Verification:
cargo test -p ailang-core.
Files this plan modifies
- Modify:
docs/DESIGN.md:2266— section header## Embedding ABI (M1)→## Embedding ABI. - Modify:
docs/DESIGN.md:2354— schema-block xref commentsee §"Embedding ABI (M1)"→see §"Embedding ABI"(inside the## Data modelwindow:2319–2558, but NOT adesign_schema_drift.rsanchor — its anchors are JSON-schema discriminator tokens, not free prose, so the drift test does not fire on this edit; recon-verified). - Modify:
crates/ailang-core/tests/docs_honesty_pin.rs:134— test comment// --- PRESENT: the DESIGN.md §"Embedding ABI (M1)" mirror ---→ drop(M1). - Modify:
crates/ailang-core/tests/docs_honesty_pin.rs:136— assert-message"DESIGN.md §Embedding ABI (M1) must mirror …"→ drop(M1). - Modify:
crates/ailang-core/specs/form_a.md:89— live forward-xrefDESIGN.md §"Embedding ABI (M1)".→DESIGN.md §"Embedding ABI".(Boss-added 5th site: a live pointer into the renamed section in a compiled-in spec — same dangling-xref class as DESIGN.md:2354; recon-confirmed NOT adocs_honesty_pin.rsasserted substring, so the edit moves no test). - DO NOT MODIFY:
crates/ailang-core/tests/docs_honesty_pin.rs:135(the asserted substring — rename-immune, stays byte-verbatim) and everyEmbedding ABI (M1)occurrence indocs/journals/**,docs/specs/**,docs/plans/**,bench/orchestrator-stats/**(committed append-only history — correctly names the section as it was at the time; renaming history is a falsification, not a tidy).
Task 1: Lockstep section rename Embedding ABI (M1) → Embedding ABI
Files:
-
Modify:
docs/DESIGN.md:2266,docs/DESIGN.md:2354 -
Modify:
crates/ailang-core/tests/docs_honesty_pin.rs:134,:136 -
Modify:
crates/ailang-core/specs/form_a.md:89 -
Step 1: Capture the baseline (live occurrences + pins green)
Run: grep -rn 'Embedding ABI (M1' docs/DESIGN.md crates/ailang-core/tests/docs_honesty_pin.rs crates/ailang-core/specs/form_a.md
Expected: exactly 5 lines —
docs/DESIGN.md:2266 (## Embedding ABI (M1)),
docs/DESIGN.md:2354 (… see §"Embedding ABI (M1)"),
crates/ailang-core/tests/docs_honesty_pin.rs:134 (comment),
crates/ailang-core/tests/docs_honesty_pin.rs:136 (assert message),
crates/ailang-core/specs/form_a.md:89 (DESIGN.md §"Embedding ABI (M1)".).
(Line 135 — the asserted substring — does NOT match this grep; it
contains no (M1) token. Confirm it is absent from the grep output.)
Run: cargo test -p ailang-core --test docs_honesty_pin
Expected: PASS — test result: ok. 5 passed; 0 failed.
Run: cargo test -p ailang-core --test design_schema_drift
Expected: PASS — test result: ok. 8 passed; 0 failed.
- Step 2: Rename the DESIGN.md section header
In docs/DESIGN.md, replace the header line (:2266):
## Embedding ABI (M1)
with:
## Embedding ABI
- Step 3: Lockstep the DESIGN.md schema-block xref comment
In docs/DESIGN.md, the schema-block line (:2354) currently reads:
"export": "<optional C symbol>", // omitted when absent (hash-stable when omitted); see §"Embedding ABI (M1)"
Replace the trailing see §"Embedding ABI (M1)" so the line reads:
"export": "<optional C symbol>", // omitted when absent (hash-stable when omitted); see §"Embedding ABI"
(This line is inside the ## Data model window :2319–2558 but the
(M1)→`` edit changes no design_schema_drift.rs anchor token —
verified Step 6.)
- Step 4: Lockstep the docs_honesty_pin test comment + assert message
In crates/ailang-core/tests/docs_honesty_pin.rs, line :134
currently reads:
// --- PRESENT: the DESIGN.md §\"Embedding ABI (M1)\" mirror ---
Replace with:
// --- PRESENT: the DESIGN.md §\"Embedding ABI\" mirror ---
And line :136 (the assert message) currently reads:
"DESIGN.md §Embedding ABI (M1) must mirror the canonical bare-scalar export-param rule");
Replace with:
"DESIGN.md §Embedding ABI must mirror the canonical bare-scalar export-param rule");
Do NOT touch line :135 (the assert!(d.contains("Export parameters are written **bare**: a scalar type carries no \own`/`borrow` mode"),` line) — the asserted substring is the
DESIGN.md body sentence, rename-immune, must stay byte-verbatim.
- Step 5: Lockstep the form_a.md forward-xref
In crates/ailang-core/specs/form_a.md, line :89 currently
contains the forward-xref token:
DESIGN.md §"Embedding ABI (M1)".
Replace that token (in place, leaving the rest of the line/sentence exactly as is) so it reads:
DESIGN.md §"Embedding ABI".
(form_a.md is compiled into ailang_core::FORM_A_SPEC via
include_str!; recon verified this line is NOT a docs_honesty_pin.rs
asserted substring — the form_a carve-out pins assert the bare-scalar
sentences at :125–132, not this xref — so the edit moves no test.)
- Step 6: Verify — zero live occurrences, pins green
Run: grep -rn 'Embedding ABI (M1' docs/DESIGN.md crates/ailang-core/tests/docs_honesty_pin.rs crates/ailang-core/specs/form_a.md
Expected: no output (exit 1) — all 5 live-coupling sites renamed;
no dangling pointer at the vanished (M1) title.
Run: cargo test -p ailang-core --test docs_honesty_pin
Expected: PASS — test result: ok. 5 passed; 0 failed. (The asserted
:135 substring is unchanged; the comment/message edits are not
asserted; the pin is green ⇒ the rename did not perturb it.)
Run: cargo test -p ailang-core --test design_schema_drift
Expected: PASS — test result: ok. 8 passed; 0 failed. (The :2354
edit changed no JSON-schema anchor token; data_model_section_is_bounded
unaffected — the section boundary ## Data model/## Pipeline is
untouched.)
- Step 7: Verify — committed history NOT falsified
Run: git diff --name-only
Expected: exactly three files —
docs/DESIGN.md, crates/ailang-core/tests/docs_honesty_pin.rs,
crates/ailang-core/specs/form_a.md. No docs/journals/**,
docs/specs/**, docs/plans/**, or bench/orchestrator-stats/**
file in the diff (those retain Embedding ABI (M1) as correct
append-only history of the section's name at the time — renaming
them would falsify the record).
Run: git diff --stat
Expected: ~5 insertions / ~5 deletions total (one line changed per
site; the header, the xref comment, the test comment, the test
message, the form_a xref). No multi-line churn.
Self-review (Step 5 — completed inline)
- Spec coverage: no spec — audit-item-driven. The audit's named
[medium](header)+[low](:2354 xref) plus the Boss-added 5th site (form_a.md:89, same dangling-xref class, recon-surfaced) and the two test-string lockstep sites (docs_honesty_pin :134/:136) are each a Step in Task 1. ✓ - Placeholder scan: no "TBD/TODO/implement later/similar to/add appropriate". ✓
- Type/name consistency: the target title
## Embedding ABIand xref form§"Embedding ABI"/§Embedding ABIare used identically across Steps 2–6 and the file-map. ✓ - Step granularity: each step is one grep / one cargo run / one one-line edit — 2–5 min. ✓
- No commit steps: none (Boss commits at iter end). ✓
- Pin/replacement substring contiguity: the ONLY presence-pin
paired with a verbatim doc edit is
docs_honesty_pin.rs:135↔DESIGN.md:2297–2299. Recon proved definitively the asserted substring (a) contains no(M1)/section-name token, (b) lives 31 lines below the renamed header, (c) is matched file-wide viad.containsnot section-scoped. Step 4 explicitly forbids touching:135and Step 2/3 edit only:2266/:2354(31+ lines away). The pinned substring is never inside any replacement body ⇒ no soft-wrap split possible. ✓ - Compile-gate vs deferred-caller: no Rust signature change;
docs_honesty_pin.rsedits are a comment + a string literal (no API, no caller cascade);form_a.mdisinclude_str!data (no recompile-cascade beyondailang-core's own test build, which Step 6 gates). N/A — no such ordering hazard. ✓ - Verification filter strings resolve:
--test docs_honesty_pinand--test design_schema_driftare recon-verified existing test targets (5 and 8 tests respectively, exact counts named); thegrep -rn 'Embedding ABI (M1'pattern is verified by recon to match exactly the 5 live sites at Step 1 and (post-edit) zero at Step 6 — a count that flips 5→0 cannot masquerade as "nothing ran" (Step 1 asserts the 5 are present first).git diff --name-onlyasserts the exact 3-file footprint so an accidental history edit cannot pass silently. ✓