Skip to content

docs: cut restated upstream behavior left by the stamp sweep (batch 5) #15513

Open

Description

@kyle-sexton

Problem

The stamp sweep in #15473 (docs: drop stamps from pointer records (batch 5)) found these passages restating upstream behavior. Under contract 6.0.0 (#15468) restated behavior is not stamped; it is cut to a pointer or rewritten as this repository's own rule. The sweep left them unchanged because cutting them changes what a skill tells the agent.

Items

  • session-flow reference/observer.md:116-118: paraphrases the precedence between CLAUDE_CODE_EFFORT_LEVEL, --effort and --settings.
  • source-control README.md:357: says subagent and workflow worktrees are kept. README.md:385: says the native binary neither ships nor uses Node.
  • source-control reference/providers/github/README.md:152-156: says the proxy serves no GraphQL (403). Same file, :177: says a draft cannot merge.
  • source-control reference/providers/github/pull-request/create.md:56-60: the proxy 403 and the REST fallback.
  • source-control reference/providers/github/babysit-prs/safety.md:47-55: gh pr merge queue behavior.
  • source-control reference/providers/github/pull-request/merge.md:169-172: what a 409 means, plus the status list. Same file, :308-311: how a PR joins the queue.
  • source-control reference/stacked-prs.md:79,85-87: stack merge semantics.
  • source-control reference/worktree-root-convention.md:83-86: where EnterWorktree(name:) lands.
  • source-control hooks/worktree-create-gate.sh:183-187: substitution-token rejection. hooks/worktree-remove-gate.sh:45-48: the WorktreeRemove contract. scripts/worktree-create.sh:1312-1320: the .worktreeinclude rules.
  • source-control skills:
    • the "files arrive through the Read tool as plain bytes" block, repeated in the babysit-loop, babysit-prs, commit, pull-request and setup SKILL.md
    • babysit-loop/reference/promotion-evidence-bootstrap.md:57-61
    • babysit-prs/reference/orchestration.md:~530
    • pull-request/reference/monitor.md:~158 and :~566
    • the claim cells of babysit-prs/reference/native-autofix-pr.md:10 and pull-request/reference/native-surfaces.md:15-19
  • planning:
    • draft-goal-condition/SKILL.md:~44: the "research preview" label
    • interview/SKILL.md:~431: Read and CLAUDE_PLUGIN_ROOT behavior
    • interview/context/loop.md:416: what the cli-reference and env-vars pages cover
  • evals:
    • design/SKILL.md:~200-205: how the arms run
    • methodology/reference/eval-design.md:~61-66: effort guidance
    • plugin-eval/reference/reading-results.md:~211-214 and the usage-limit Gotchas bullet in plugin-eval/SKILL.md: errored-run grading
    • methodology/reference/local-decisions.md:74-75: the blog's rubric position

Kept, unsure

  • planning skills/draft-goal-condition/reference/workflow-row.md:19,35,47: kept, as ruled. The skill names this record's date when it cannot fetch the docs, and the dg-standing-setting-xhigh eval samples cite it.
  • evals "distilled 2026-08-08" stamps in methodology/SKILL.md:19, eval-design.md:10, grading.md:10, recipes.md:9 and success-criteria.md:9: kept, as ruled. The update action reads and refreshes them.
  • source-control skills/babysit-loop/reference/promotion-evidence-bootstrap.md (~428, ~444) and promotion-evidence-implementation-plan.md:36: verification records of our own code, not of an upstream.
  • source-control skills/worktree/reference/gather-block.md:42-47: an owner record that mixes a probe, the 2.1.257 note and pointers.
  • planning lib/html-escape.mjs:21: a generated copy. Its "Recheck:" pointer comment belongs to the canonical lib/html-escape.mjs, which batch 7 owns.

Kept, no anchor checked

  • evals skills/design/SKILL.md (old 209): plugin-evals.md has no anchor.
  • planning skills/design/reference/type-discipline/typescript.md:129 (Handbook Narrowing) and interview/context/nfr-prompts.md:39 (ISO page).
  • session-flow reference/observer.md:87: the cli-reference link. It is a stored value, so its stamp stays.
  • source-control:
    • README.md:385 (/docs/en/setup)
    • reference/providers/github/README.md:65-67 and :178
    • reference/providers/github/pull-request/ready-for-review.md:~246
    • skills/babysit-loop/reference/promotion-evidence-bootstrap.md:42
    • skills/pull-request/reference/create.md:425
    • skills/babysit-prs/reference/stuck-checks.md:76
    • skills/babysit-prs/reference/native-autofix-pr.md:10

Do

Per item: cut to the anchored pointer, keep as our own rule with its reason, or record why it stays. A cut that changes agent behavior needs an eval or a stated basis.

Related

Refs: #15473, #15468

Activity

  1. added
    needs-triageNot yet classified. Floor until a type and one priority tier are set.
    on Nov 12, 2026
  2. github-actions commented on Nov 12, 2026

    @github-actions
    Contributor

    Size: Large
    Complexity: Medium
    Uncertainty: Knowable
    Flags: none
    Blocked by: none

    Understood: this is the batch 5 follow-up to the stamp sweep in #15473 under contract 6.0.0 (#15468). Passages that restate upstream behavior in the session-flow, source-control, planning and evals plugins each get cut to an anchored pointer, kept as this repository's own rule with its reason, or recorded as staying. Cuts that change agent behavior need an eval or a stated basis.

    The issue is docs-only, so I gave it low priority. It spans four plugins and many files, so it may need splitting. I found no duplicate or merged fix. The similar open issues cover other batches (6, 7, docs/, playbooks and others), not these items.

    Automated triage. A maintainer reviews it.

  3. removed
    needs-triageNot yet classified. Floor until a type and one priority tier are set.
    on Nov 12, 2026
  4. melodic-automation-lanes commented on Nov 12, 2026

    @melodic-automation-lanes
    Contributor

    Not admitted to the issue lane: no trusted-human acceptance criteria: a listed human must put the confirmed - [ ] criteria in the body or post them as their own comment, or, on an issue a listed human filed, intake's newest refinement comment must hold them unedited. Admission takes the issue once each of those is met, when intake triages it again (.github/actions/resolve-config/admission-rules.json).

  5. melodic-automation-lanes commented on Nov 12, 2026

    @melodic-automation-lanes
    Contributor

    Refinement. A read-only run explored this repository, searched its issues and pull requests, and read the allowed docs to draft what follows. On an issue a listed human filed, it counts as the issue's spec once posted, while unedited; it is still model text, so check it.

    Intended outcome: Each listed passage across session-flow, source-control, planning and evals is cut to an anchored pointer, kept as this repository's own rule with its reason, or recorded as staying, with an eval or stated basis for any cut that changes agent behavior.

    Related work

    Open questions

    1. For each of the ~35 items, is the decision cut to pointer, keep as own rule, or keep and record why? The issue leaves that choice to the run.
      RECOMMENDED: Owner lists a per-item verdict, or authorizes cut-to-pointer as the default for pure upstream restatements (proxy 403, queue behavior, 409 meaning, EnterWorktree landing, effort precedence) and keep-as-own-rule where a hook or script enforces it (worktree-create-gate.sh, worktree-remove-gate.sh, worktree-create.sh). Basis: judgment
    2. Which cuts change what a skill tells the agent, and what basis is accepted for them: a new eval, or a stated rationale?
      RECOMMENDED: Accept a stated basis for pure-docs files (READMEs, reference records). Require an eval run only for SKILL.md instruction changes such as the repeated Read-tool plain-bytes block in five skills. Basis: judgment
    3. The effort-precedence paragraph in observer.md:116-118 sits beside a probed 2.1.296 finding and the dual drop-and-pin workaround at :108-114. Should the Record be cut to the three doc pointers, or kept because it justifies this repository's own rule?
      RECOMMENDED: Keep the workaround as the repository's own rule with its probe, and cut only the Claim restatement at :116-118 to the existing anchors. Basis: plugins/session-flow/reference/observer.md:108-122
    4. The "Kept, no anchor checked" items have no verified upstream anchor. Should this run verify anchors on code.claude.com, or leave them out of scope?
      RECOMMENDED: Out of scope. Record them as staying until an anchor is checked. Basis: judgment
    5. The work spans four plugins and about 35 files. Should it be split per plugin into separate issues?
      RECOMMENDED: Split into four issues (session-flow plus planning, source-control README and providers, source-control skills and hooks, evals), each with a task list of its items. Basis: judgment

    Sources read

    • /home/runner/work/claude-code-plugins/claude-code-plugins/refine-data/context.json: issue 15513 lists about 35 items with no task list, no per-item verdict and no comments from the owner.
    • plugins/session-flow/reference/observer.md:116-122: the Record restates the effort precedence and already cites env-vars#precedence, settings#settings-precedence and model-config#adjust-effort-level with an As of: 2026-10-10 stamp.

    To answer, post a comment that starts with this line:

    **Decision recorded.** <answer>
    

    Automated refinement by the lanes App. Answer with a comment whose first line is **Decision recorded.**.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    Fields

    Work status

    needs-owner-decision

    Priority

    low

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions