docs(upstream): record the unslop lane in the cursor/plugins attribution table

[kyle-sexton]

No linked issue.

Summary

docs/upstream/cursor-pstack.md calls itself the "Single source of truth for everything in this
marketplace derived from cursor/plugins … and specifically its pstack/skills/ collection", and its
·r·echeck trigger is "a change to any pstack/skills/<name>/SKILL.md named in the attribution table
below".

unslop is a member of that collection, and plugins/ai-slop is derived from it (#3031, ai-slop
0.2.0) — but the table had no unslop row. So the single-source-of-truth claim was not true, and
upstream drift in pstack/skills/unslop/SKILL.md would have gone undetected, because the trigger
only reaches rows the table names. Before this change the file mentioned ai-slop or unslop only
in passing, inside the technical-writing row.

One row added. Nothing else changed.

Fix

An unslop row in the attribution table, in the same shape and voice as the existing eight, with
Relation: Absorbed (skill omitted) — ai-slop:audit already owned the axis over a
Wikipedia-derived tell inventory and already shipped the detect-then-guarded-fix flow, so the port
landed as entries in an existing catalog rather than as a new skill.

The row records what was taken (three script rules, four rubric tells, the plain-word vocabulary
trio, and the fix-time half that became reference/rewrite-guide.md), what was deduplicated against
the Wikipedia inventory via the catalog's overlap map, and what was rejected with its reason —
Name-dropping as out of scope for general prose, the general half of Generic conclusions, abstract
metaphor nouns as a script rule (calibration measured "substrate" in 114 legitimate technical uses on
this corpus), and one adding-soul bullet.

It also records what it is not. The row states plainly that this verdict was formed at
integration time against upstream main with no revision recorded — the catalog and the rewrite
guide both cite an unpinned blob/main URL, and the catalog's own four-part drift record covers the
Wikipedia source page only. So the file's main@60c641e4 pin is this row's baseline for the next
diff, not the state it was audited at. It also dates the derivation explicitly (ai-slop 0.2.0,
2026-08-19) rather than letting this file's git history imply the row's own creation date.

This head is a conflicted merge, resolved by hand — please check that first

The branch was opened 29 commits behind and conflicted in the very file it edits. #3069 had
rewritten the arena, technical-writing and blast-radius rows on main, including a paragraph
in arena — "Adaptation notes, corrected on verification" — that retracts two earlier overstatements
about upstream's isolation: frontmatter and its judge instruction.

Resolved by taking main's three rewritten rows verbatim and appending only the new row. Taking
the branch's stale side would have silently reverted someone else's correction, which is the worst
outcome available here and the thing worth reviewing first.

Verified rather than asserted, against the merge's own second parent:

$ git diff --numstat e5015213..HEAD          # whole repo
1	0	docs/upstream/cursor-pstack.md          # one insertion, zero deletions

per-row sha256, arena / technical-writing / blast-radius:
  e5015213  b74b44c3d4d995a2  16f99c48a1801206  4de853aa1196780f
  HEAD      b74b44c3d4d995a2  16f99c48a1801206  4de853aa1196780f

#3069's correction text survives verbatim at HEAD. Note that silent-revert-canary.yml runs
post-merge and never inspects a PR, so nothing automated would have caught a revert here.

Verification

Every factual claim in the row was checked against the repo, and the upstream-side claims against
cursor/plugins at the pin:

Check	Result
All 12 named rule ids exist in reference/catalog.md	present
Tier claims vs the detector-findings crosswalk	rule-chatbot-artifacts IMPORTANT; rule-filler-phrases, rule-stacked-hedging SUGGESTION
The four rubric tells are rubric, not script	all four v1: rubric; zero hits in detect.sh
The three "dormant" rules are dormant in both layers	all v1: recorded-only; absent from detect.sh's registry, and the rubric layer is defined as the v1: rubric entries
utilize / leverage / facilitate in the shipped default vocabulary	present in DEFAULT_VOCAB
Substitution guardrail as quoted	"Never parentheses, never en dashes, never a spaced hyphen"
The 114-uses calibration figure	catalog.md and ai-slop's CHANGELOG
Overlap map says "catalogued by" rather than "covered by"	present
Overlap-map completeness	upstream ships 31 numbered patterns; 8 absorbed by the 7 new rules, 23 accounted for by map rows (8 + 23 = 31)
Upstream-side spot checks at the pin	description ends "Must always apply"; six adding-soul bullets, five survive; #20 and #22 separate upstream, merged here
Pin disclaimer accurate in both directions	no revision pin exists anywhere in plugins/ai-slop/ — all five citations are blob/main, so the row neither overstates nor understates
Table integrity	9 contiguous rows, 4 columns, Relation byte-identical to sibling usage
"All ten lanes are decided" still holds	ten = 8 table rows + bro + show-me-your-work, which have decision sections without rows
markdownlint-cli2, typos, editorconfig	clean

Related

• feat(ai-slop): unslop-inspired catalog additions, plus the corrections and dogfood config that make them true (0.2.0) #3031 — the ai-slop 0.2.0 integration this row documents.
• feat: port the cursor/plugins pstack collection — 2 new skills, 8 argued absorb/omit verdicts #3065 — added docs/upstream/cursor-pstack.md and its attribution table.
• fix: land the pstack port's post-merge verification findings across eight plugins #3069 — rewrote three rows on main; its corrections are preserved through this merge.

Known remaining gap, not closed here

show-me-your-work is also a pstack/skills/ member with shipped derivatives in this marketplace —
plugins/implementation 0.15.0 (four rules absorbed into implement-dispatch's DEVIATIONS.md) and
plugins/claude-ops (the formula-injection guard) — and it likewise has no attribution row, so the
·r·echeck trigger does not reach it either. That is a pre-existing second instance of the same gap,
neither created nor worsened here; it is tracked as a follow-up. bro correctly needs no row, having
absorbed nothing.

[claude]

Claude finished @kyle-sexton's task in 2m 17s —— View job

---

Code review complete

• Gather context (PR diff, CLAUDE.md, code-review skill criteria)
• Apply skip gate — no skip; single-row table addition to a "source of truth" doc has real review surface (factual accuracy of every claim)
• Review changed file(s) against criteria
• Post findings

Scope

Single file, +1/-0: docs/upstream/cursor-pstack.md — one new row added to the attribution table for unslop.

What I checked

This lane reviews for correctness and alignment with repo conventions. Since the change is prose making factual claims about other files in the repo, I independently re-ran the PR's own verification table against the current tree rather than trusting the author's claims:

Claim	Independently verified
All 12 rule ids (rule-chatbot-artifacts, rule-filler-phrases, rule-stacked-hedging, rule-false-ranges, rule-colon-crutch, rule-abstract-metaphor-jargon, rule-mechanism-free-claims, rule-canned-notability, rule-challenges-conclusion, rule-bold-overuse, rule-inline-header-lists, rule-title-case) exist in reference/catalog.md	✅ confirmed present
Tiers in the crosswalk: rule-chatbot-artifacts → IMPORTANT; rule-filler-phrases, rule-stacked-hedging → SUGGESTION	✅ matches docs/conventions/detector-findings/README.md:231-233 exactly
utilize / leverage / facilitate in the shipped default AI-vocabulary	✅ present in scripts/detect.sh's DEFAULT_VOCAB
Substitution-guardrail quote ("never parentheses, never en dashes, never a spaced hyphen")	✅ verbatim in reference/rewrite-guide.md:17-18
The 114-uses calibration figure for "substrate"	✅ present in both catalog.md:94 and CHANGELOG.md:198
Overlap map wording "catalogued by" (not "covered by")	✅ confirmed at catalog.md:699
Table integrity — row/column count	✅ 9 data rows total, every row (header through the new unslop row) has exactly 5 `
Every relative link in the new row (catalog.md, rewrite-guide.md, CHANGELOG.md, detector-findings/README.md)	✅ all four paths exist
ai-slop 0.2.0 referenced in the row exists in the changelog	✅ CHANGELOG.md:188 ## [0.2.0]
No pre-existing unslop/ai-slop mentions elsewhere in the file conflict with the new row	✅ only the pre-existing technical-writing row and the new row reference it, no contradiction

I did not independently re-run markdownlint-cli2 (not available in this sandbox without network access), so that specific line of the author's verification table is author-claimed and unverified by me — everything else in the table above I re-derived from the repo directly rather than restating the author's claim.

Findings

None. Every factual claim in the new row checks out against the current tree, the row's column count and format match the other eight rows exactly, and the "not audited at the pin above" caveat is consistent with the file's actual main@60c641e4 pin (docs/upstream/cursor-pstack.md:9) and the trigger it clarifies (docs/upstream/cursor-pstack.md:12-14). No correctness, maintainability, or convention issues found.
· branch claude/pstack-unslop-attribution-row

[github-actions]

Claude has reviewed this PR 2 times. The lane skips further automatic reviews after 5; deleting this comment resets the count.

[claude]

Claude finished @kyle-sexton's task in 3m 14s —— View job

---

CI code review complete

• Gather context (PR diff, CLAUDE.md, /review:code-review skill criteria)
• Apply skip gate — no skip; single-row table addition to a "source of truth" doc has real review surface (factual accuracy of every claim)
• Review changed file(s) against criteria
• Post findings

Scope

Single file, +1/-0 vs origin/main: docs/upstream/cursor-pstack.md. This head (785537c) is one commit past the head this lane already reviewed (10f74e9), so this pass re-checks the whole row and focuses verification on what that last commit actually changed.

What changed since the last review on this PR

The last commit ("date the unslop derivation, not the row") inserted one clause into the row:

— derived at ai-slop 0.2.0 on 2026-08-19 (reference/catalog.md, "Second pass"), so this file's git history dates the row, not the derivation —

I independently checked this addition against the repo rather than trusting it:

Claim	Independently verified
reference/catalog.md records a "Second pass" dated 2026-08-19 for the Cursor additions	✅ plugins/ai-slop/skills/audit/reference/catalog.md:87 — "Second pass, 2026-08-19, for the Cursor additions, against the same corpus:"
The row's own commit history postdates that derivation (so citing git history as the derivation date would be wrong)	✅ git log --date=short on this file: the row was added/amended 2026-08-21 (526f02e, 785537c), three days after the 2026-08-19 derivation date it now cites explicitly instead
"not one of the ten lanes decided below" is consistent with the file's own count	✅ the pre-existing ## Not adopted section's closing line ("All ten lanes are decided.", cursor-pstack.md:372) is unchanged by this PR and counts the table's 8 pre-existing rows plus bro and show-me-your-work (decided in prose only) = 10; the new unslop row is explicitly an 11th, out-of-band addition, and doesn't require bumping that count since it says so itself

All of the other claims in the row (rule ids, tiers, vocabulary, the 114-uses figure, the overlap-map wording, link targets) are unchanged from the version already verified in this lane's prior pass on 10f74e9, and remain correct.

I did not independently re-run markdownlint-cli2 — not available in this sandbox (network access to fetch/run it is blocked here) — so that line of the author's own verification table is author-claimed and unverified by me.

Findings

None. The added clause is accurate, well-cited, and resolves a real gap the author flagged in the PR description (the file's git-history date and the actual derivation date would otherwise have been conflated). No correctness, maintainability, or convention issues found.
· branch claude/pstack-unslop-attribution-row