Add repository review and PR skills and refresh engineering context

[dustin-sale]

Description

Add repository-local local-review and pr-authoring skills, and make engineering guidance easier to discover and apply. Local review identifies committed and dirty snapshots, selects native validation routes, and reports evidence limits. PR authoring uses the selected base's template and supports scoped draft creation.

• Add guides for code quality, test quality, FastMCP compatibility, pagination and tool safety, with concise root/shared routes and owning quality requirements.
• Replace the unsafe pagination loop in the best-practices example with a signature-only illustration; explain remaining-allowance requests, hard output caps, backend adapters, continuation and concrete regression cases.
• Clarify successful-result redaction, hidden effects, overwrite preconditions, fail-closed authentication and submission-versus-observation behavior. Recommend clearly synthetic password/secret values in unit tests.
• Refresh Common consumer/dependency context, Recovery authentication/validation guidance and Data Studio's documented coverage threshold against current source.
• Record the context assessment, scoped design, implementation and evaluation evidence. Issue authoring remains deferred.

No application code, tests, dependencies or runtime response shapes change. No new dependencies or dependent releases are required. Implementation gaps described by the guidance remain separate follow-up work.

Tracking issue: omitted with the author's explicit waiver for this draft PR.

Type of change

• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

How Has This Been Tested?

Documentation-appropriate checks run from the repository root against committed HEAD a6ff06cc7d7079b7271a1f23ee3702db9c0576e1:

• All 21 changed Markdown files checked: 710 local links and 156 anchors resolved, including root/shared discovery routes; relevant source references inspected.
• Both repository skills passed the installed skill-creator frontmatter validator using the existing virtual environment.
• git diff --check 3a7cfd9623ec97ad7c365b9b77f28e459aefca8c...HEAD passed.

To reproduce: resolve file-relative links and heading anchors in the changed Markdown files; compare documented source/validation references with their owning files; validate both .agents/skills/ folders with skill-creator's scripts/quick_validate.py; run the diff check above. No dependency installation is needed for the prose checks.

The implementation record documents six review and seven PR-authoring scenarios, repeated controlled trials, and an independent review of the initial skill/guidance snapshot. Those synthetic text-only trials did not create real PRs or run server tests. Subsequent pagination, test-secret and tool-safety edits received focused source/link/scope/diff checks; the earlier review is not presented as a fresh review of this complete committed branch.

Application suites, coverage, live OCI/database operations and deployed authorization checks were not run for this documentation/skills change. Structural validation does not prove runtime correctness or future agent performance.

Test Configuration:

• Firmware version: Not applicable.
• Hardware: Not applicable; no hardware-specific behavior changed.
• Toolchain: Git, temporary Python documentation checker, existing repository virtual environment and installed skill-creator validator.
• SDK: No SDK operations executed; source/manifests/locks were inspected for documentation accuracy.

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• Any dependent changes have been merged and published in downstream modules

Unchecked personal attestations are not inferred from agent review. Code/test checklist items are not applicable to runtime behavior here; application tests were not run, and there are no dependent releases.