feat(stability): add race-condition-repro skill - #97
Conversation
…rency The falsifier for a concurrency claim is a test that never interleaved: a sequential run exercises no race and produces a green indistinguishable from a real pass. The skill therefore treats showing the interleaving occurred as the proof obligation, not the assertion passing. Lands in `stability/` beside `memory-leak-hunt`, the other defect-class engine `pr-validate` delegates to.
MajorLift
changed the title
race-condition-proof skill
…xing it The description read "the engine behind evidence's B7 deterministic-interleaving evidence category". "B7" is an address into evidence-catalog.md, not a name: it carries no meaning to a reader who has not opened the catalog, and a frontmatter description cannot link out to one. - Drop the lane id from the description; name the role instead. - Replace the trailing Notes reference with a Related section that links the catalog by URL. A relative path would not survive installation, which flattens skills to mms-<name>/. - Drop the exogram-daemon path, which is a private repo the reader cannot open. - Update the stale pr-validate name to evidence.
As a noun suffix, `-proof` is privative in English — waterproof, bulletproof, tamper-proof all mean "immune to". So `race-condition-proof` parses as "immune to race conditions" rather than "produces a proof about ordering", and that misreading is plausible enough not to self-correct. `-repro` names what the harness produces and carries no such inversion.
MajorLift
changed the title
race-condition-proof skillrace-condition-repro skill
Context budgetWhat this PR costs an agent, measured from an install rather than read from the diff. Three tiers, and only the first is unavoidable.
Frontmatter is the only tier paid unconditionally — every agent loads it on every run once the skill is installed, used or not, because it is what the agent reads to decide relevance. The 28 skills across the eleven open skill PRs sit at a median of ~1,716 tokens selected and ~1,860 with references followed. All are within the 1,536-character description budget. Selected is paid only when the agent picks the skill. + refs & knowledge is the ceiling if every bundled reference is then read; it is a worst case, not an expectation. Method
These figures are pinned to the commit above and drift on every push; #96 tracks automating them. |
1 participant
Summary
race-condition-repro, the engine behindevidence's B7 deterministic-interleaving evidence category and the siblingred-on-basenames for ordering bugs.domains/stability/besidememory-leak, the other defect-class engineevidencedelegates to.Test plan
node .github/scripts/lint-skill-entry.mjs domains/stability/skills/race-condition-repro/skill.md— 0 errorsnamematches directory,maturity: experimental, description within the 1536-char budgetdomains/stability/is also created by feat(stability): addmemory-leakskill #81; domains are discovered viareaddirSync, not a registry, so both can add it independentlyValidation runs
Trial runs of this PR's skills against merged
metamask-extensionPRs nobody flagged. Every claim was re-verified against the real diff before posting. Clean results are included on purpose — a skill that only ever reports problems cannot be calibrated.race-condition-reprorace-condition-reproEach comment carries a trial-run disclaimer and links back here for feedback.