docs(skills): preserve RPaaS audience across lint migrations

[Yuchao Yan (msyyc)]

Summary

Make subsequent RPaaS migrations smoother without broadening RPaaS restrictions to ordinary ARM or weakening validation/recovery gates. This is a standalone, skill-only follow-up to the first ExtensionResourcePathPattern migration in #5682 and promotion in #5684.

Process findings and changes

Observed evidence	Skill correction
Intake initially stopped because no automatic native RPaaS selector was established. The accepted continuation used existing explicit lint enablement, default-disabled, with ordinary ARM counterexamples. Upstream RPaaS registration had been flattened to catalog common/Both.	Establish upstream audience separately from catalog eligibility. Consider supported native selection, an independently justified broader guideline, or documented external opt-in before classifying. Record mixed-audience limitations; preserve genuine uncertainty stops.
The full generic corpus selected 468 projects, compiled 462, and left six failures. Target 0/0 reflected a default-disabled rule and ordinary ARM selection, not RPaaS parity.	Carry actual validator/native activation profiles into migration notes, promotion, and both reviews. Require enabled focused evidence and appropriate activation integration controls; retain required representative/full regression runs and disclose audience, version, diagnostic-unit, and source-identity limits.
Development review reproduced two warnings for one authored operation-template endpoint, including a template companion. Four focused regressions corrected the diagnostic unit.	Check rule-owned HTTP/template traversal and exact endpoint counts/targets before expensive corpus runs, while retaining the exclusion of framework-only tests.
Promotion documentation regeneration exited successfully but did not load the authored rule Markdown until supported compiler docs/fileRef metadata was added.	Verify authored-file binding and generated rationale/examples, not merely a successful regeneration or index entry.
A quiet successful lint command produced no Tee-Object receipt file. Guessed config paths and a stale readiness budget mirror also caused avoidable handoff failures.	Capture empty raw output independently, derive paths from the actual checkout, and acknowledge authoritative later scoped grants without inventing stale-mirror prerequisites. Preserve exit statuses, counters, fingerprints, and failed evidence.

The reusable audience contract lives in rule intake; development, queue handoffs, and promotion link to it rather than duplicating a new RPaaS detector. Promotion also checks library/default/recommended/all activation, including automatic all behavior, and preserves every existing rule's enablement.

Evidence and boundaries

• The reviewed source's migration evidence documents the native opt-in contract, ordinary ARM scope distinction, template regression, corpus limitations, and intentional diagnostic-unit/version differences.
• The original stop, accepted continuation, failed-command receipts, scoped grants, and handoffs remain in the coordinating session's durable process records. Those operational artifacts are not committed.
• Existing capabilities/reviewer ownership, immutable-source review, finite recovery authorization, and no-valid-comments termination already cover the runtime/review portions of the run and remain unchanged. The earlier optional .git repository-identity clarification remains separate in Clarify canonical Git remote URL identity #5685.
• No rule/package files, completed migration commits, later promotion changes, catalog schema, new worker flags, production audience detectors, dependencies, release entries, or generated outputs are changed. This PR does not claim production RPaaS corpus parity or re-review subsequent rule-PR heads.

Validation

• Reviewed the complete six-file Markdown diff against fetched canonical feature/lintdiff-migration-new base 527a585925e61bd782e61cce9127e235e555ea91, including linked native-test responsibility and recovery/ownership contracts.
• Ran Prettier write/check on only the six explicit changed files using already-prepared repository tooling; all passed.
• Verified 76 relative Markdown file references and all new audience-contract anchors.
• git diff --check passed.
• No dependency installation, submodule setup, package/corpus builds/tests, repository-wide formatting, or recursive review loop was needed for this documentation-only update.