agent: builder model: opus max_turns: 100 budget_usd: 26 —
yaml: explicit keys, directives, limitsThe owner’s direction: an official package yaml with a zero-copy tree pointing into the input, spans, safety limits on by default (alias expansion, depth, size), YAML 1.2 core schema, tags kept, a duplicate key an error, << not merged by itself. Every gap in the compiler, lowering or std is a finding fixed in the compiler, never routed around in the package.
Read first: docs/specification/adr/adr-yaml-1.md (D2-D5 and D8 for this brief), then packages/yaml/src/main.nika and its tests as y2-tree-properties.md (block scalars) and y2c-properties.md (properties, aliases; reports beside them) left them; y1’s report too. Its Facts the architect probed hold here too: input ref Bytes, rule WS = "" -> { }, the tree built inside the grammar’s actions (#1000), entries as a struct (#999). Also read y5-escapes-and-radix.md: its Facts the last builder paid for hold here, and the Scalar shape it left (a double-quoted scalar’s value decoded by the grammar, every other scalar’s value a view, ADR-yaml-1 D4 amended, D11).
The check happens once, where the knowledge is (owner, 2026-10-30; ADR-yaml-1 D5’s last paragraph, D11): what the grammar can check (a block scalar header, a tag’s shape, a %YAML version) it checks with a guard, so a mistake is a parse error with line and column, and no later function checks it again.
Props, TagKind handed to the scalar rules, fits_node): read its commit message; its tags !e!x are kept as written, this brief resolves them.y17-1041-entry-takes-a-parameter.md, the brief before this one) lets parse(input: ref Bytes, limits: Limits) hand limits to the entry rule; before step 2 run its test tests/compiler/tests/grammar_an_entry_takes_a_parameter_of_its_caller.nika. If it is not on develop or fails, stop after step 1 and report partial with Needs: #1041; do not pass the fields one by one around it.String input, scratch file): a rule taking ds: Vec[Dir] (a struct of two views made by an earlier rule dir -> Dir, %TAG !e! tag:ex,2000:) and an i64 read in a guard, called as t:Tag(ds, lim.depth) after ds:dir*, prints tag:ex,2000:x for !e!x; an entry rule doc(lim: Limits) given a struct literal or a local works. So the %TAG handles can be handed down the rules as a parameter, and the depth as an i64 parameter checked by a guard.return (rustc today, #1040 waits on the owner): compute the value with a let mut and a loop.Part: #928 in the report, never Fixes: none (y2c’s commit message read Fixes none).nikaia fmt keeps a grammar rule’s guard whole now (#1015, 2e553fe5); run it on every changed .nika before the gate.? key / : value in block and flow, a collection as a key; %YAML 1.2 (another major version an error, 1.1 accepted as 1.2), %TAG !e! tag:example.com,2000: applied to !e!x in that document only; a directive needs ---. Tests tests/explicit_keys.nika, tests/directives.nika.pub struct Limits { pub depth: i64, pub alias_nodes: i64, pub bytes: i64 } with the defaults of D8 (alias_nodes is checked by load, a later brief: store it, document it), and parse taking a Limits the caller may pass (omitted: the defaults; follow how an entry rule takes values after its input, tests/compiler/tests/grammar_an_entry_takes_values_after_its_input.nika). The parse refuses nesting deeper than depth (a depth parameter on the rules and a guard) and an input longer than bytes, with an Error naming the limit, its value and the position. Test tests/limits_depth_and_size.nika: 129 nested flow sequences refused by default, accepted with depth: 200; 128 accepted; an input over a small bytes refused.The expected .stdout is what YAML 1.2.2 says; the yaml-test-suite case for a construct is the reference (read only, nothing copied: ADR-yaml-1 D9 waits for the owner); name it in a comment.
target/fast/nikaia test --project packages/yaml passes (the earlier tests included); the package gate below is green; each step committed and pushed; the last commit’s message says Part of #928.
docs/builder-facts.md first. develop in the main checkout; git status at start: other agents’ uncommitted files are not yours, never touch or stage them.git add <files> only; commit after each step; git push origin develop fast-forward, no rebase (rejected: report Pushed: rejected, go on committing).cargo fmt, cargo clippy --all-targets -- -D warnings, scripts/pre-push-check.sh, and the gate of .claude/agents/builder.md (#852).CHANGELOG.md, badges; no new NK code. New tests in Nikaia only; never the full suite. No foreground sleep, no polling; scripts in files.std or the toolchain lacks stops the work: no code around it unless an ADR or Part I-III refuses it (name it). Else Result: blocked, Found: gap + probe, Needs: owner decision (CLAUDE.md, A missing form stops the work).NIKAIA_CACHE_LIMIT=2G; at most two clones with target/ in the machine; delete your own clone’s target/ and target/release/nikaia-scratch at the end; report GB freed.Pushed: line names the pushed sha. A report without a Pushed: line, or with a placeholder left in (TC, DISK), is not done and cuts the batch (k3a1 2026-10-29: Result: done, nothing committed).NIKAIA_CACHE_LIMIT=2G target/fast/nikaia test --project packages/yaml green, and cargo test --profile fast -p nikaia --test it -- fmt::every_nika_file (it formats every .nika under packages/). For a change only under packages/yaml/ this is the gate the line above asks for; a change under crates/ runs the gate of .claude/agents/builder.md (#852) as well. target/fast/nikaia is built by the session hook; if it is missing, cargo build --profile fast -p nikaia once.Changelog: none, unless the brief changes crates/.Files not to touch: anything outside packages/yaml/ and the brief’s own report, unless the brief names it; never docs/briefs/batch.txt or a brief of another lane.
max_turns: commit and push every finished step, then report Result: partial with what is left under Needs:. Uncommitted work at the end is lost work (y4 ran out at 110)./home/user/Nikaia/docs/briefs/yaml/y2b-keys-directives-limits.md.report.md)Fields of docs/architect.md, Briefs and reports: Result: done|noop|partial|blocked (the bare word alone on that line: Result: done, never a summary after it, which cuts the batch, b869a 2026-10-27; Owner: is a question for the owner or none, never a role), Files, Commit, Body (optional, one paragraph for the commit message), Fixes, Pushed, Tests, Gate (green|red), Deviation, Found, Needs, Changelog, Owner, Disk freed, and for a move Rs2nika: (the kept N of M lines of rs2nika --score, then the three costliest hand edits; none for a task that moves no Rust, none (redesign) for a brief with Shape: redesign and no Convert: line). none where nothing, and then the bare word: Deviation: none (a note) cuts the batch (batch.nika compares exactly); a note goes into the commit message. Found: is only for a defect outside your change (it cuts the batch); a process event such as another session resetting or rewriting your files is no Found:, it goes under Needs: (m3a 2026-10-28); a converter draft that came out wrong is no Found:, it is one of the costliest edits under Rs2nika:, which the converter lane reads (e12-4b 2026-10-27); how you solved your own step goes into the commit message. Deviation: is only a step of the brief built otherwise than written or not built (a choice the brief left to you, such as keep what X needs and say so, is no deviation: it goes into the commit message; nor is building two steps in one commit when every step is built, e12-3a 2026-10-27, nor a step another commit on develop already built, shown by your test failing with that commit’s change reverted, b889 2026-10-27: name the commit in Body:, nor a test’s expected .stdout/.stderr taken from what the program really prints where the brief’s sketch of it was short or wrong, b934 2026-10-29: say so in Body:, nor a lowering that follows the ADR governing it where the brief’s expected lowering contradicts that ADR, b933 2026-10-29: name the ADR in Body:, nor a lowering other than the brief’s where rustc cannot take the brief’s, b1029 2026-10-31: say why in Body:, nor re-keying an allow-list entry (KNOWN) whose hash changed for the same program and the same reason, b-fixforward-sweep2 2026-10-28: say so in Body:, nor an existing test’s expected .stdout/.stderr updated because your change rightly changes the message it prints, b1012b 2026-10-31: name the test in Body:); how you verified (e.g. a test not run against the unfixed compiler) goes under Tests:. A noop says why under Tests: (what was probed), not under Deviation:. Final answer, one line: done <sha>, report <path> (or BLOCKED: ... / PUSH REJECTED ...).
Attribution: Co-Authored-By: Claude Opus 5.5 noreply@anthropic.com Claude-Session: https://claude.ai/code/session_01UVxJDveRM5qKZKyCqq6Y24
A first builder on this brief was stopped mid-work when its driver hit a time limit. Its unfinished work is saved outside the checkout: /tmp/claude-0/-home-user-Nikaia/e431aae9-4983-5166-8583-f9b3eb3b6f61/scratchpad/y2b-partial/tracked.diff (the changes to tracked files, including packages/yaml/src/main.nika) and untracked.tar in the same folder (new tests: directives, explicit_keys, limits_depth_and_size). Read it, take over what holds up, and check every piece of it against this brief. It has not passed any gate.