Skip to content

moooon

Motir

Vibe your whole project. Bring an idea — Motir's three AI layers plan it, track it, and ship it, end to end. You're looking at Motir, built in Motir.

  • Vibe Project
  • Open Source
  • AI Agent
  • AI Loop
1
requests
0
upvotes
145
planned
1,361
shipped

Motir · Work items

MOTIR-2660Done

(motir-meta) Compress the rest of `core.md` — the planning rules and the plan-time gates

Repo: motir-meta. One PR. The second execution card under MOTIR-2642's COMPRESSION.md, applying the mechanics MOTIR-2659 built.

MOTIR-2659 compresses core.md's per-card gate checklist (lines 31–337). This card takes the remaining 392 linesPlanning rules (line 338 onward: recommend-the-best, the decision-authority ladder, plan-with-the-coding-conventions, cardinality) and The plan-time gates (line 695 onward).

core.md is the only pack every pass pays for, in every phase, on every kind: 699 lines, 24.5% of the loadable corpus, 75.1% of the cheapest cell. It is the one file where COMPRESSION.md § Decision 3's cap binds, and the cap cannot be met by MOTIR-2659 alone.

Acceptance criteria

  1. Every warrant span removed from core.md appears verbatim in prompts/plan-rules/fixtures/core.md, under the anchor its inline tag points at — appended to the file MOTIR-2659 created, not a second file.
  2. Each compressed rule carries exactly one evidence tag in COMPRESSION.md § Decision 1's four-slot form.
  3. No rule's TRIGGER or DIRECTIVE is re-worded. The deletion test (§ Decision 1) is applied span by span; a span whose removal changes when the rule fires stays inline, and a rule that cannot be compressed without touching its trigger is left whole and named in the PR body.
  4. AMENDED 2026-08-11, on the record — see the amendment note below. python3 prompts/plan-rules/COMPRESSION.measure.py --check is run and its core.md number is reported in the PR body with the deletion-test enumeration behind it. Where AC 3 and the 520-line cap conflict, AC 3 wins — § Decision 4: "If a pack cannot reach its number without cutting a trigger, a tell-list or a shape table, the target gives way, not the rule." A miss is not a failure of this card; it is the measurement § Decision 3's "raising the cap is a decision record, not a commit — it amends this one, with a measurement" asks for, and it is owed to the amendment card.
  5. WARRANT CONSERVATION is proved in the PR body (invariant C): the diff contains only deletions whose text appears verbatim in fixtures/core.md, plus inserted tag lines matching the fixed form. The proof is COMPRESSION.conserve.py's WORD-level check run against the pre-compression tree (--base 5b32d96), not a line-level read of the diff — the tool's own header says a cut re-wraps the lines around it and a line-level diff cannot tell that from a re-wording.
  6. DECISION.md, CLASSIFICATION.md and plan-rules.md are not edited.

⚠️ Amendment note — AC 4 (2026-08-11)

As authored, AC 4 read: "python3 prompts/plan-rules/COMPRESSION.measure.py --check reports core.md at or under 520 lines (§ Decision 3's cap), down from 699."

It was not reachable, and the evidence is a count rather than a judgement. The run delivered 36 warrant spans (670 → 598; 699 pre-compression) with invariant C green. Closing the last 78 lines requires cutting triggers, tell-lists or the index, which AC 3 forbids.

520 came from the measure script's projection of 501, and that projection's warrant detector is a regex for citation markers. Enumerated on the tree as it stands, 22 marker-bearing sentences remain in core.md outside the tags and not one is warrant under the deletion test: 9 are rule HEADINGS whose attribution is part of the rule's name, 6 are TELLS, 7 are DIRECTIVES carrying a pointer or a scope clause. A further ~38 line-equivalents are the evidence tags this compression itself inserted — they start with >, carry a key and a date, and score as compressible warrant. The projection is self-referential and drifts upward as warrant is removed: 501 before MOTIR-2659, 542 after it, 546 mid-way through this card.

§ Decision 1 sanctions the outcome in advance — "A pack that misses its number because of them has found something worth knowing." The cap amendment is MOTIR-2706; the projection defect is MOTIR-2707.

Context refs

  • prompts/plan-rules/COMPRESSION.md — § Decision 1 (the tag form + the deletion test), § Decision 3 (the 520-line cap and what a new rule pays), invariant C.
  • prompts/plan-rules/core.md lines 338–699.
  • prompts/plan-rules/fixtures/core.md — created by MOTIR-2659; this card appends to it.