Skip to content

docs(knowledge): align model-matching table with dateless model-ID scheme - #1928

Merged
kyle-sexton merged 1 commit into
mainfrom
docs/row228-model-ids
Aug 4, 2026
Merged

docs(knowledge): align model-matching table with dateless model-ID scheme#1928
kyle-sexton merged 1 commit into
mainfrom
docs/row228-model-ids

Conversation

@kyle-sexton

Copy link
Copy Markdown
Contributor

No linked issue.

Summary

Doc-alignment campaign row 228 ("Model IDs and versioning",
https://platform.claude.com/docs/en/about-claude/models/model-ids-and-versions).

The Anthropic profile's digest-agent model-matching table pinned model-specific guides to
"a full model ID — never a bare family alias, which resolves to the current family model".
The live page now documents that since the 4.6 generation the canonical model ID is dateless
(claude-{name}-{major}[-{minor}]) and "is not an alias. It is the snapshot" — so the old
wording would misclassify exactly the correct pin for a current-generation model guide as a
forbidden alias and wrongly fall through to the session default.

Changes

  • Reword the table cell to "its pinned model ID — never an alias that can move to a newer
    snapshot", dropping the stale dated-era vocabulary.
  • Add one sentence under the table routing the generation-dependent pinned-vs-alias
    resolution to the live page at spawn time (pointer-not-copy — no model-ID tables copied).
  • Bump knowledge plugin 0.10.21 → 0.10.22 with changelog entry.

Evidence

  • Live fetch 2026-08-04, raw-markdown channel (.md appended), 3836 bytes / 83 lines,
    channel re-verified working for this doc.
  • Page: "This guarantee covers model IDs, not the convenience aliases that the Claude API
    accepts for some earlier models" (pinning guarantee); "A 4.6-generation ID such as
    claude-sonnet-4-6 is not an alias. It is the snapshot" (drift trigger).

Related

  • Doc-alignment close-out campaign roster row 228 (.work/doc-alignment/ROSTER.md).

🤖 Generated with Claude Code

…heme

The Anthropic profile's digest-agent model-matching table pinned model-
specific guides to "a full model ID — never a bare family alias, which
resolves to the current family model". The live model-IDs-and-versioning
page now documents that since the 4.6 generation the canonical model ID
is dateless ("is not an alias. It is the snapshot"), so the old wording
would misclassify the correct pin for a current-generation model guide
as a forbidden alias and wrongly fall through to the session default.

Reword the cell to "its pinned model ID — never an alias that can move
to a newer snapshot" and route the generation-dependent pinned-vs-alias
resolution to the live page at spawn time (pointer-not-copy). Verified
against the live page 2026-08-04 via the raw .md channel (3836 bytes).

Bump knowledge plugin to 0.10.22 with changelog entry.

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
@cursor

cursor Bot commented Aug 4, 2026

Copy link
Copy Markdown

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

@claude

claude Bot commented Aug 4, 2026

Copy link
Copy Markdown

Claude finished @kyle-sexton's task in 1m 21s —— View job


Review: docs(knowledge) — align model-matching table with dateless model-ID scheme

  • Gather context (diff, changed files)
  • Understand the request
  • Review changed files
  • Post findings

Scope reviewed: plugins/knowledge/.claude-plugin/plugin.json, plugins/knowledge/CHANGELOG.md, plugins/knowledge/skills/docpage-digest/context/anthropic-docs-profile.md (per git diff origin/main...HEAD).

Summary

Straightforward, well-scoped doc-alignment change. No code logic involved — this is a prose/table wording fix plus a pointer sentence to the live model-IDs page. I don't have findings that need fixing.

Notes (no action required)

  • Version bump correctness: 0.10.210.10.22 patch bump in plugin.json is consistent with the CHANGELOG entry and CLAUDE.md's "version is the delivery vehicle" rule — good.
  • Pointer-not-copy discipline: the new sentence in anthropic-docs-profile.md#L179-L182 correctly routes generation-dependent resolution to the live page rather than embedding a model-ID table snapshot in the profile — matches this repo's stated anti-staleness pattern and avoids re-introducing the exact drift this PR is fixing.
  • No duplicate references: confirmed the model-ids-and-versions URL only appears once in the file, so this isn't creating a second inconsistent copy of the same guidance elsewhere.
  • Table cell wording (line 175) reads correctly — antecedent of "which would digest a historical guide on the wrong version" resolves unambiguously to "an alias," and the "When no pinned ID is resolvable, omit the override" fallback still matches the row's original intent.
  • Fresh-docs mandate: PR body cites a live fetch (2026-08-04, raw-markdown channel, 3836 bytes) with two direct quotes from the source page as evidence for the "dateless ID is the snapshot" claim — satisfies the sourcing requirement for this class of change. I can't independently re-fetch the page in this review, so I'm trusting the cited evidence as presented.
  • Out of scope for this lane: no security-relevant surface (no manifest/hook/MCP contract change, no code execution) — nothing to hand off to the security lane.

@claude

claude Bot commented Aug 4, 2026

Copy link
Copy Markdown

Claude finished @kyle-sexton's task in 1m 18s —— View job


Security Review — PR #1928

  • Fetch PR diff / read changed files
  • Assess for logic flaws, injection, secrets handling, workflow permission changes
  • Post findings

Scope: plugins/knowledge/.claude-plugin/plugin.json (version bump 0.10.21 → 0.10.22), plugins/knowledge/CHANGELOG.md (+15 prose lines), plugins/knowledge/skills/docpage-digest/context/anthropic-docs-profile.md (+6/-1, table-cell reword plus one routing sentence).

(Note: git fetch/gh pr diff were blocked by sandbox approval in this run, so I read the three changed files in full on the PR branch directly — diffs are small enough that full-file review gives complete coverage of what changed.)

Findings

No security issues found.

This is a docs-only change:

  • No workflow files (.github/workflows/**) are touched, so no permissions:, trigger, or pull_request_target/workflow_run changes are in scope.
  • plugin.json's only change is the version field bump — no new userConfig keys, no permission or capability widening, no new MCP/tool wiring.
  • CHANGELOG.md is inert prose.
  • anthropic-docs-profile.md is a static, maintainer-authored context file bundled with the plugin (not fetched from an untrusted runtime source), consumed as instructions by the docpage-digest skill's agent. The edit rewords one table cell and adds a sentence pointing to a first-party platform.claude.com URL — no user-controlled input, no new external fetch target, no secrets/tokens/credentials touched, no shell/script/template content introduced.

Nothing here creates a new injection surface, alters trust boundaries, or changes what code executes.

@github-actions

github-actions Bot commented Aug 4, 2026

Copy link
Copy Markdown

Claude has reviewed this PR 1 time. The lane skips further automatic reviews after 5; deleting this comment resets the count.

@kyle-sexton
kyle-sexton merged commit 8642374 into main Aug 4, 2026
32 checks passed
@kyle-sexton
kyle-sexton deleted the docs/row228-model-ids branch August 4, 2026 16:49
kyle-sexton added a commit that referenced this pull request Aug 5, 2026
…#1936)

## Summary

Implements the PA-H interview decision (locked 2026-08-02) in the
Anthropic docs profile: the
applicability filter gains one `tag-exempt (<sub-shape>)` disposition
for rows carrying no
guidance for ANY surface the vocabulary adjudicates, with four
documented sub-shapes
(consumer-surface, archive-descriptive, metadata, navigation-pointer),
the sub-shape named at the
row. The class describes the material's genre and asserts nothing about
harness applicability -
not a positive tag, not a negative claim - so it owes no live-doc
citation and no absence basis,
and the near-miss disclosure burden never attaches. `api-only` remains
reserved for rows that DO
assert a harness absence for their own specific assertion. Consistent
with the co-decided PA-AI
Reading 1 (a positive tag asserts harness applicability and requires a
live-doc citation).

- `anthropic-docs-profile.md`: vocabulary intro mentions the tag-exempt
disposition; new bullet in
  the applicability-filter section defines the class
- knowledge plugin `0.10.22` -> `0.10.23` + CHANGELOG entry

The memory-tier retag sweep over the system-prompts slice (the
decision's implementation cost) is
performed separately in `.work/` and is never committed.

No linked issue

## Related

- PR #1898 (archive-reading conventions, knowledge 0.10.20) - same
profile, same campaign
- PR #1928 (model-matching table alignment, knowledge 0.10.22) - most
recent edit to this profile

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant