docs(knowledge): add the tag-exempt class to the applicability filter

[kyle-sexton]

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 docs: ship DOC bundle 4 — matrix pointers, archive-reading conventions, residual-risk footer #1898 (archive-reading conventions, knowledge 0.10.20) - same profile, same campaign
• PR docs(knowledge): align model-matching table with dateless model-ID scheme #1928 (model-matching table alignment, knowledge 0.10.22) - most recent edit to this profile

🤖 Generated with Claude Code

[cursor]

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 finished @kyle-sexton's task in 1m 4s —— View job

---

Security review

• Gather context (PR diff via local commit 27f2e6d vs parent 297a925)
• Review each changed file for security issues
• Report findings

Scope reviewed (all 3 changed files, +30/-3):

• plugins/knowledge/.claude-plugin/plugin.json — version bump 0.10.22 → 0.10.23 only
• plugins/knowledge/CHANGELOG.md — new [0.10.23] entry (prose)
• plugins/knowledge/skills/docpage-digest/context/anthropic-docs-profile.md — new tag-exempt (<sub-shape>) disposition documented in the applicability-filter section

Findings: none.

This PR is a pure documentation/policy change to a skill's context file, plus a manifest version bump and matching changelog entry. There is no code execution path here to review:

• No scripts, hooks, .mcp.json, or GitHub Actions workflow files are touched.
• No new tool permissions, allowed-tools, or permissions: blocks are introduced or widened.
• plugin.json's only change is the version string — no new fields, no new trust delegation.
• The markdown content is static prose defining a documentation-tagging vocabulary (tag-exempt, its four sub-shapes) for a downstream docpage-digest skill to apply when reading Anthropic docs. It contains no URLs, no embedded commands, no templated/interpolated content, and nothing that would be interpreted as instructions to an agent beyond ordinary skill-context prose (i.e., no prompt-injection vectors introduced).
• Nothing in the diff touches secrets, credentials, tokens, authorization logic, or data egress.

Given the diff is entirely non-executable documentation plus a version/changelog bump, this PR is out of scope for the categories called out (injection surfaces, workflow trigger risk, permission widening, supply-chain pin risk) — there simply isn't a corresponding surface changed here.

[claude]

Claude finished @kyle-sexton's task in 2m 26s —— View job

---

Code review

• Gather context (PR diff, changed files)
• Review anthropic-docs-profile.md changes for correctness and consistency
• Review CHANGELOG.md entry and version bump
• Post findings

Scope reviewed (per REVIEW.md lane split — security is out of scope here, already covered by the security lane above):

• plugins/knowledge/skills/docpage-digest/context/anthropic-docs-profile.md
• plugins/knowledge/CHANGELOG.md
• plugins/knowledge/.claude-plugin/plugin.json

Findings

No blocking issues. This is a well-scoped, docs-only addition:

• New tag-exempt (<sub-shape>) bullet is internally consistent with the surrounding filter. It correctly stays orthogonal to api-only's absence-basis/near-miss machinery (lines 83–111): the near-miss disclosure rule is scoped to falsifying api-only, so a disposition that asserts nothing about harness applicability legitimately owes neither a live-doc citation nor a near-miss note. The "api-only remains reserved for rows that DO assert a harness absence" sentence (L135–136) closes the boundary explicitly, which prevents the two dispositions from being confused at authoring time.
• Intro sentence update (L73–76) correctly threads the new disposition into the tag enumeration without disturbing the existing cc-applicable/mixed vs. api-only bullet ordering (a pre-existing minor ordering inconsistency between the intro list and the bullets below — not introduced by this PR, not worth fixing here).
• CHANGELOG.md entry matches the profile change precisely and follows the file's established per-entry style (compare to the 0.10.20/0.10.21/0.10.22 entries) — good adherence to this repo's documentation conventions.
• Version bump 0.10.22 → 0.10.23 is consistent with this plugin's pattern of a patch bump per doc-content addition, and per CLAUDE.md's versioning rule, this is required since the profile file ships to consumers only when the manifest version increases.
• Four sub-shapes (consumer-surface, archive-descriptive, metadata, navigation-pointer) are collectively exhaustive enough as documented, and archive-descriptive ties back cleanly to the existing "Archive-reading conventions" section (L44–69) rather than inventing an unrelated concept.
• No test coverage applicable — this is prose-only skill context consumed by an LLM at digest time, not executable code, so there's no unit/integration surface to add coverage to.

One non-blocking note: I couldn't run markdownlint-cli2 against the changed files in this sandbox (no network access for npx, and the dependency isn't vendored in node_modules) to mechanically confirm MD004/MD007 list-style compliance. Manual inspection shows the new bullet uses the same dash-bullet/backtick/bold conventions as the rest of the file, so this is low-risk, but flagging in case CI's lint step should be the final word.

[github-actions]

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