Repository navigation
Closed
Labels
agent-readyFully specified and briefed; eligible for autonomous pickup from the frontier.Fully specified and briefed; eligible for autonomous pickup from the frontier.
Description
Parent
Refs #15703
Source: PLAN Phase 14 (first work item), topic ecosystem-specifics
Depends on
Outcome
The classification rules for model-read ecosystem specifics are shown to decide the same way for two independent readers, the rows to cut are counted per plugin, and one cut item per plugin is filed in the planned order.
Demo: the classification record on this item shows zero split rules (or the owner's ruling on each), and the per-plugin cut items exist under the container with their row counts.
Steps:
- Blind classification: two fresh agents each classify a stratified sample of about 50 inventory rows (at least 5 per kind) under the design's classification rules R1 to R11, without seeing the other's answers.
- A rule is split when they disagree on any sampled row that either placed under it. Reword a split rule and re-classify the disagreeing rows plus ten fresh rows, at most three rounds. A rule still split after three rounds goes to the owner; label this item
needs-humanat that point. - Count the rows to cut per plugin and state the number on this item.
- File one cut item per plugin under the container, in this order: the discovery explore fallback table; the review ecosystem-specialist agent's generic defaults; mutation-testing's per-tool tables; user-interface's framework pages with their concerns index and its tests; planning's TypeScript idioms page with the type-discipline contract's sentence; then every other plugin with rows under rule R11, the plugin with the most rows first. A shipped default that picks a library goes with its plugin's cut.
- Each cut item says: list every item with its rule; back every R5 or R11 call with a search showing whether a script reads the file; freeze content-dependent cases first; carry the sensitivity-control result for each whole-page deletion and the detection-control result for each partial cut; run no more than 400 runs per plugin (the changed skills' cases on the
opusandsonnetaliases, three runs per side, plus control sides on one model). - Each cut item blocks the gate-widening item; add those edges when filing.
Decisions
- Q27: a classification rule is split when the two blind classifiers disagree on any sampled row that either placed under it; at most three rounds of rewording, then the owner decides.
- Q28: the remaining cuts go in the design's named order, then by plugin with the most rows to cut first.
- D3: each plugin's cut is its own topic and pull request.
- Interview Q32 (D4 of the Brief): bundled defaults that pick a tool, command or library are retired wherever they appear.
- Bulk-mutation standard: count first, then act; each pull request states its row count.
Key interfaces
- "The consumer's ecosystem (language, framework, tools): abstract now. Seam: the ecosystem-resolution contract and the consumer's own records."
- Conventions followed: a cut ships with eval evidence (the skills-and-agents review rules).
Done when
- The classification record shows zero split rules, or names the owner's ruling on each.
- The per-plugin row counts are stated on this item.
- One cut item per plugin exists under the container, each blocking the gate-widening item, and their number is stated before filing.
- AC1 and AC7 are carried into each cut item's Done when (no generic skill ships a model-read tool, library or command choice outside the named exceptions; each whole-page deletion carries a sensitivity-control result and each partial cut a detection-control result).
- AC10: no new enforcement-placing skill is added here.
Out of scope
- Cutting any guidance (the per-plugin items do that).
- Widening the portability gate.
Metadata
Metadata
Assignees
Labels
agent-readyFully specified and briefed; eligible for autonomous pickup from the frontier.Fully specified and briefed; eligible for autonomous pickup from the frontier.
Type
Fields
Work status
confirmed
Priority
medium
Activity
github-actions commentedon Nov 13, 2026
Size: Large
Complexity: Medium
Uncertainty: Knowable
Flags: none
Blocked by: none
Blast radius: tracker records only (a classification record, per-plugin row counts, new cut items under #15703); no guidance is cut here.
Reversible: yes; the counts and filed items can be edited or closed.
Open questions: none
Decision recorded by the latest owner comment:
That decision settles the round-4 open questions. The remaining steps (full pass over the refreshed rows, per-plugin counts, filing cut items) are host work already routed via
host-run. No other open issue or merged PR duplicates this request.Automated triage. A maintainer reviews it.
melodic-automation-lanes commentedon Nov 14, 2026
Not admitted to the issue lane: it carries host-run. Admission takes the issue once each of those is met, when intake triages it again (
.github/actions/resolve-config/admission-rules.json).melodic-automation-lanes commentedon Nov 15, 2026
{"v":1,"actor_id":337595936,"title_sha256":"9d79b74de8166e30535bbccff3f3fc28f2e061133b75b36cc237ef795d7c3808","body_sha256":"9828144936142295b079e265a2acb47574f048184c5eeae6c50ed72b8b726179"}
github-actions commentedon Nov 15, 2026
melodic-automation-lanes commentedon Nov 15, 2026
The work-resolve-issue lane stopped on this issue (
host-only): the run reports the work needs a host run the lane cannot make.It added
host-runand set Work status to confirmed: an agent run on a host builds it, and it waits on no decision from a person.In the run's words:
What the run says it did not do:
Run: https://github.com/melodic-software/claude-code-plugins/actions/runs/43770041352
To hand it back to the lane, remove
host-runand set Work status to agent-ready again.2 remaining items
kyle-sexton commentedon Nov 16, 2026
Addendum to the record: R6 and R7 agree only on a shared non-literal reading. Both round 1 readers reported, independently, that the rule order read literally lets R6 (mechanism record) match every model-read contract id first, so R7 (contract value) would never apply. Each then placed those rows under R7 anyway, with R6 as runner-up, which is why the comparison shows no split there. A reader who follows first-match exactly would place them under R6. The dispositions differ (R6 keeps one copy; R7 renames ids that name a tool for a mechanism class), so this is not a label-only question. R6 and R7 were not reworded in rounds 2 and 3. Those rounds' fresh rows held one model-read contract id (reserve row V13, round 3), and both readers placed it under R3, an earlier rule, so the R6 against R7 order was not tested again.
For the owner, with the R3 question above: state in R6 that an ecosystem-specific id other skills or config write or match is judged by R7, not R6 (recommended; it is the reading both readers chose), or keep first-match as written and drop R7's model-read case.
kyle-sexton commentedon Nov 16, 2026
Decision recorded.
Owner, 2026-11-16, asked whether a skill that lists the tools or formats it supports in its description thereby puts its model-read guidance for those tools in declared scope (rule R3), with the recommendation "no": "go w/ recommendations" (said in the hub session, relayed to this one).
needs-humanuntil those are answered.kyle-sexton commentedon Nov 16, 2026
Decision recorded (AC1 exception set), with what stays open.
Read against the texts in a fresh context, then each cited line re-read:
docs/adr/0101-resolve-ecosystem-specifics-from-the-consumer-repository.md:49).Not settled by the texts, left for the owner (
needs-humanstays):For each: either the ADR gains a kept class for it, or the rule changes from stays to goes. The full pass does not restart until that, the count scope and the R6 against R7 order are answered.
kyle-sexton commentedon Nov 17, 2026
Classification record: round 4 (rules revision 4), 2026-11-17
Posted by an agent session. Nothing is filed and nothing is counted yet: round 4 split, so the
full pass, the per-plugin table and the cut-item count wait on the owner's answers below.
needs-humanstays on this item.What changed since the last record
"ADR 0101 gains kept classes for R6's third clause and R8; the count covers R11 plus failing R4
and R8; R6 defers ecosystem ids to R7"), the owner's R3 ruling (a description that lists tools
does not make declared scope), and the "No ecosystem-specific examples" rule merged in PR 16634 /
16649:
names; a description listing the formats a general capability supports does not qualify.
default" dropped.
classified.
mainat 5b66a7f (by script): 528 model-read rows outside K9/K10plus 25 script- or both-read K3 rows (an R4 check, since a script-read value can still pick a
library) = 553. 17 rows are gone because every cited file is gone on
main(15 + 2: architecture1, biome-format 3, debugging 1, harness-ops 1, ruff-format 5, testing 3 (the property pages),
typos-format 3). 536 rows remain (513 + 23). Readers also answer
GONEwhen a cited passage wasrewritten away in a file that still exists. The design's R8 example row (
finding-classes.md:51)no longer holds its pairing on
main, so round 4 used two other R8 candidates.Round 4
Two fresh readers, given only the rules and the rows, classified 36 rows: the round-1 R3
disagreements, rows the widened revision-3 R3 caught, eight model-read K6 rows (two whose values a
script of ours may emit), five K7 rows, a tool-doc link row, two R8 candidates, three R4 rows and
seven fresh controls. Compared by script.
map-flowbody (K4)The readings, quoted:
skill descriptions name RPG Maker MZ as an output target." B: "R3 fails because every pixel-art
description lists MZ beside other targets (Godot, PICO-8, plain strips, 9-slice, blob-47), which
is a general capability listing the targets it supports."
adapter reads only tracked
*.csfiles and refuses any other language." B: "C# is the onelanguage a general entry-point tracing capability supports, which R3 excludes."
covers." B: "The cited rung ids are this marketplace's ladder ids, and its scripts emit them:
rules-ledger.mjs:50printsCHECKING_RUNGS...,rules-ledger.mjs:143andrule-inventory.sh:214printsemgrep-rulein their old-id notices."semgrep-rulerung id as an alias; these are model-read K6 idsthat skills match, and no script emits them." B: "its rung ids are values
rules-ledger.mjsprints ... and names in its old-id message."
the mechanism it describes." B: "The manifest description states a behavior of this marketplace
... and no other skill or config writes or matches that text, so the R6 hand-off to R7 does not
apply."
Per this item's rule, a split after the owner's decisions goes to the owner rather than to another
rewording round.
Owner question 1: R3 for one named target or one-language reader
When a skill's description names a platform or language and the row's content serves only it, but
the capability could serve others (pixel-art lists MZ beside Godot and PICO-8;
map-flowtracesC# only), is the row declared scope under R3?
and the row's content is an artifact for, or a reader of, only that named one; a description
that lists several formats one general reader handles is not declared scope. Basis: the design
states that R3 extends the philosophy's inherently-locked test to output targets and fixed reader
formats, which revision 4 keeps in its examples ("a game engine layout", "a skill whose product is
reading those formats and whose shipped reader parses only them").
is that platform's artifact) is declared scope; per-target content of a multi-target skill is
judged by the later rules. Closer to "genuinely and inherently locked"
(
docs/plugin-philosophy.md:458); F0316 becomes a cut (R11).Owner question 2: what "a value one of its scripts emits" covers in R6
Do ladder rung ids that a script holds in a list or names in a deprecation notice, and a manifest's
coverage claim, go to R6 or R7? Both keep the row; R7's disposition renames a tool-named rung id
and opens an ecosystem-id list, R6's keeps the text once.
(a report field such as
cov_source), not an id a script validates against or names in a notice.Basis: R7's disposition names both cases outright ("a ladder rung ... is renamed"; "a coverage
claim changes with the mechanism it describes"), and the design limits R7 to mechanism-class tool
ids and ecosystem ids as data.
prints.
Answering both unblocks: the full pass over the 536 refreshed rows, the per-plugin table, the
cut-item count, and removal of
needs-human.Correction to this item's text
model)" is superseded by the Resolve ecosystem specifics from the consumer repository and align the enforcement ladder with the talk's pyramid #15703 decision of 2026-11-16
(issuecomment-6906800940:
no paired no-loss compare per cut) and the owner's session decision of one targeted case per cut,
sonnet, 3 runs a side.Overlap noted for the review cut item
An unpushed branch for #16646 (no PR yet) edits
plugins/review/agents/ecosystem-specialist.md:it adds a prerequisite paragraph right after the stack-default fallback lines (
dotnet test,pytest,ruff checkand the rest) that the review cut item covers, and a new report examplenaming one ecosystem's tools (a YAML lint skip naming
ajvandnpm ci), which revision 4 wouldplace under R11. The review cut item should be drafted against that branch's text once it merges.
Cut items that would be filed: not yet counted (blocked by the two questions above). None is filed.
kyle-sexton commentedon Nov 17, 2026
Decision recorded. The owner answered both round-4 questions on 2026-11-17, typed in a session: "go w/ recommendations" to each, with the options restated in full.
Rules revision 5 is revision 4 with R3, R6 and R7 worded to these two answers; every other rule is unchanged. Round 4's five split rows under revision 5:
map-flowbody*.csBoth readers' answers on each row are now reconcilable: the readers split on what the two questions asked, and no row stays open. Only F0316 changes the cut count (it stays). Next: the full pass over the 536 refreshed rows under revision 5, then the per-plugin table and the cut-item count. Nothing is filed.
kyle-sexton commentedon Nov 17, 2026
Per-plugin cut count (done-when 2 and 3)
Posted by an agent session. Full pass under rules revision 5 over the 536 refreshed rows (input at
5b66a7fa), one reader per chunk, tallied by script (first answer per row). 536 of 536 answered, no conflicts.Per-rule totals over all rows: {'R1': 2, 'R2': 31, 'R3': 75, 'R4': 24, 'R5': 12, 'R6': 87, 'R7': 23, 'R8': 7, 'R9': 6, 'R11': 257, 'GONE': 12}.
Rows to cut (R11 + R4 + R8): 288 in total (R11 257, R4 24, R8 7). Toolchain's 33 are excluded because #15733 owns that cut, leaving 255 rows across 47 plugins. R8 rows are cut unless their targeted case fails without them.
47 cut items will be filed under #15703, each blocking #15732, in this order: 1. discovery, 2. review, 3. mutation-testing, 4. user-interface, 5. planning, 6. testing, 7. code-tidying, 8. tdd, 9. docs-hygiene, 10. implementation, 11. source-control, 12. harness-config, 13. instruction-placement, 14. knowledge, 15. performance, 16. verification, 17. work-items, 18. app-control, 19. code-metrics, 20. debugging, 21. harness-memory, 22. improvement, 23. ai-slop, 24. developer-experience, 25. education, 26. mcp-tools, 27. prototype, 28. session-flow, 29. architecture, 30. codebase-health, 31. corrections, 32. disk-hygiene, 33. github, 34. playbooks, 35. adhd, 36. ai-briefing, 37. bugs, 38. coupling, 39. dometrain, 40. evals, 41. event-storming, 42. machine-health, 43. playwright, 44. skill-quality, 45. songwriting, 46. visualization, 47. wizard.
The issue's named cuts come first (discovery explore fallback table; review ecosystem-specialist; mutation-testing per-tool tables; user-interface framework pages; planning TypeScript idioms page), then the rest by rows to cut, most first.
The eval recipe in this issue's Steps ("no more than 400 runs per plugin ...
opusandsonnetaliases, three runs per side") is superseded for every cut item: one targeted before-and-after case per changed skill onsonnetat 3 runs a side,opusonly to confirm a sonnet loss, no paired compare, each whole-page cut keeping its own sensitivity control (#15703 comment 6906800940; #16672 comment 6935228111). Each cut item states it.Cut rows with two readings
A reader's note named another rule for these cut rows; each cut item marks the row and quotes the note. In 36 of 44 the other rule is one that keeps rows (some notes name it only for part of the cited range).
Approval
The owner approved filing on 2026-11-17, asked "File the 47 cut items for #15730?", option A "file all 47, in the tally's order"; answer verbatim: "go w/ recommendations".
kyle-sexton commentedon Nov 17, 2026
The done-when is met; closing. Posted by an agent session.
agent-readyand Work statusready. The owner approved filing all 47 in the tally's order on 2026-11-17 ("go w/ recommendations" to that option).Stated gaps:
kyle-sexton commentedon Nov 19, 2026
Report from cut item #16764 (pull request #17470), under its rule "A row whose file a script reads is not cut here: report it on #15730."
Row F0212, file
plugins/harness-config/skills/audit/reference/required-permissions.md: a script reads this file.plugins/harness-config/skills/audit/scripts/audit-engine.shuses it as its baseline, taking only table rows whose first cell is a backtickedTool(...)pattern under thesensitive-file-deny,destructive-bash-denyandask-rulesheadings.Pull request #17470 still changed three prose passages in that file (the
python -c/node -eexamples, "a Python or Node script", and theBash(npm run *)wildcard example). No table row changed, so the script's input is the same before and after. The cut was kept on that basis; the pull request's eval has no case for this file.Posted by an agent session.