Repository navigation
docs: drop stamps from pointer records (batch 7) - #15476
kyle-sexton wants to merge 4 commits into
Conversation
Apply upstream-drift contract 5.0.0 to the plugins no other batch owns and the root lib/ and scripts/ canonicals: pointer records lose their as-of date and recheck trigger, stored values keep theirs, and recorded decisions keep a dateless named trigger. Shared copies are re-synced, and every touched plugin gets a patch fragment. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VcupojQwC9uJAfmSt4EHZm
Per owner decision D26, pointer rows in Claim/Basis/As of/Recheck tables show "—" in the As of and Recheck cells; value rows keep their date. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VcupojQwC9uJAfmSt4EHZm
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c13177969f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
claude-security-review partial review of Security review: PR #15476 (docs: drop stamps from pointer records, batch 7)I found no security issues. This was a full review, not blocking only: this lane has not reviewed an earlier head of this PR. critical-findings: 0 Coverage. I read every hunk of the 319 distinct changed files in What I checked:
|
Per D26, a record holding a version, count, byte size, hash, timing or exit code read from upstream keeps its as-of date and recheck trigger. Restore 33 such stamps the sweep removed, including the Mermaid probe counts in visualization's decision matrix. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VcupojQwC9uJAfmSt4EHZm
|
claude-review partial review of Code review: PR #15476, drop stamps from pointer records (batch 7)Result: no findings. I read the whole diff: all 319 distinct file changes, plus the 237 identical copies listed in What changed. Comments and docs only. The PR removes What I checked:
Settled thread. I did not re-raise the Other inputs:
Not verified. I had no shell, so I ran no tests or hygiene gates. The PR body's before/after stamp counts were provided by the author; I did not check them. |
The em dash in pointer rows of mixed As-of tables fails the purged-em-dashes check; leave those cells empty and reword the intros. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VcupojQwC9uJAfmSt4EHZm
|
Held: 556 changed files exceed what the review lanes read whole. The fable-5-2 effort session splits it into smaller batch PRs and closes this one as superseded; nothing here merges as-is. |
No related issue: batch 7 of the Q11 sweep that applies the pointer-without-stamps policy from #15449; it has no tracking issue.
Summary
This PR applies upstream-drift contract 5.0.0 ("Required parts") to the 71 plugins no other batch owns, plus the canonical files in root
lib/andscripts/. The shared copies were re-synced from those canonicals, so copies inside other batches' plugins change too, and every touched plugin gets one patch fragment. CHANGELOG.md,evals/data,vendor/, YAML/JSON stores and generated files were not touched.[shell command execution disabled by policy]in the setup skills), measured figures, hashes, pinned SHAs and code comments beside constants.Fix
Counts are stamp lines (
**As of**,As of:,Recheck trigger,recheck when, inlineas of YYYY-MM-DD). They were counted with a script that skips shared-copy files, CHANGELOG.md,evals/,vendor/and.changes/. Base is0545c99c1e4.The lines that remain are stored values, dateless named triggers, prose that only mentions the phrase "recheck trigger", and the kept-unsure items below.
Other edits:
discipline/evals/swa-fork-off-degrades,developer-experience/evals/review-app-start-handles,multi-agent/evals/as-disabled-switch-downgradeandmulti-agent/evals/su-apply-guard-off-confirms.routines.md,guardrail-slice.md, instruction-placementsources.md,remove-shims.sh, writingsources.mdand developer-experiencecli-contract.md.Restated behavior to cut (79)
CLAUDE_PLUGIN_DATAand another plugin's hook can export onecommand the cmdline NUL layoutConfigChangefires after any write and cannot asknodeon PATH and how a hook that cannot start is treated$.session.usage()reports and where mods runused_percentagecounts the input side onlyCLAUDE_PLUGIN_ROOTRegisterdoc comment inindex.d.ts.commandreaches Claude as a notification" restates the monitors section.maxEffortLevelbehaves.data:and AVIF pages (repeats claude-code#79824 and #81410).CLAUDE_CODE_MAX_TOOL_USE_CONCURRENCYcaps, and the subagent limit.${CLAUDE_SKILL_DIR}substitution..claude/commandsstatus.-preads stdin as the prompt.inPlace: truewriteseleven_v4PostToolBatch" restates the hooks page/timingretirement, the per-search cap, and thatactions_listhas nocreatedfilterGet-Secretsearch and-AsPlainTextbehaviorpass/gpg-agent promptingDependency "<name>" is disabledKept, unsure (53)
gated: false)UIDconstant (Node's Windows chmod behavior); kept under the constant-comment rule.gatingHooksandhasCatchare upstream field names the script reads.check-skill.shflag the#Nas a history citation, so the stamp stays until that check accepts a dateless trigger.-shellcheck=decision[shell command execution disabled by policy]word for word. About 18 setup skills carry the same record, so the batch should decide it the same way everywhere.nodeLinkerpointer. The text lists the upstream value names (pnp,node-modules,pnpm), so I treated them as stored strings.mcp__plugin_<plugin>_github__<tool>name form, a string copied from upstream.~/.claude/plugins/data/<id>/, a string copied from upstream.CLAUDE_CODE_SESSION_ATTENDEDvalue (2.1.301) and thepermissionDecision/askstrings the script writes.FINDINGconstant, which names an env var, and holds an open question (feat(guardrails): evaluate CLAUDE_CODE_OVERLOADED_RETRY_MAX_DELAY_MS in workflow-resilience-check #6756) with its own trigger.sg). They record option sets the guard's model depends on.--disallow-toolshelp-line observation (2.1.316).falsefor an absentdisable-model-invocation, kept as a stored default.mechanicalmodel-default record. I kept the full stamp as a model default; it could also be read as a recorded decision with a named trigger.plugins/discovery/skills/report/SKILL.md(scripts/validate-plugin-contracts.mjs:1162-1186;validate-plugin-contracts.test.sh:1219). Remove the stamps from the canonical file and then from every copy together._metaargument and attribute names.batched-pass.md:30-32cites it as the file's opening stamp.Claim: as of 2026-09-29comments about our own code, not about an upstream page.effort: highconstant.run_attempt,run_started_at,created_at) that the commands readmarkdownlint-cli2 --helpoutput (banner first, no--version), observed tool behaviorKept, no anchor checked (36)
SLASH_COMMAND_TOOL_CHAR_BUDGET(stored-value record, kept stamp)https://code.claude.com/docs/en/permissionswith no anchor. The likely anchor is#what-read-and-edit-rules-cover, which guardrails setup SKILL.md already uses.https://git-scm.com/docs/git-worktree.permissionspage, no anchor.skillspage, no anchor.context7.com/docs(stored-value record)Restored value stamps
Per D26, a record that holds a version, count, byte size, hash, timing or exit code read from upstream keeps its stamp. The third commit restores 33 stamps (in 23 files) that the sweep had removed: the visualization decision-matrix Mermaid probe counts, plus version-bearing records in autonomy, context-budget, context-guard, explainer-video, firecrawl, fleet, instruction-placement, mcp-tools, playwright, rate-limit-guard, skill-quality and toolchain. The other 61 regex candidates turned out to be behavior-only or noise (IPs, exhibit and arXiv ids, our own constants). These were left removed:
Withheld
skills/report/SKILL.mdin implementation and plugin-quality must stay byte-identical toplugins/discovery/skills/report/SKILL.md(scripts/validate-plugin-contracts.mjs). Discovery belongs to another batch, so these copies keep their stamps here.Verification
markdownlint-cli2on the 276 changed .md files: 0 issues.lychee --offlineon the 463 changed files: 0 errors.check-skill.shon each scope plugin's skills root: 388 WARN before and 381 after. No plugin went up, and none failed.sync-shared-copies.sh --check: all 802 copies match.check-changelog-fragments.sh --check: all 95 fragments are valid. Each of the 92 touched plugins has one patch fragment.instruction-placementremove-shims.test.sh:257, which asserts that the record has no**As of**, still holds.Related
🤖 Generated with Claude Code
https://claude.ai/code/session_01VcupojQwC9uJAfmSt4EHZm