@@ -1,7 +1,7 @@
 # Developer Instructions
 
-**Version**: 9.27
-**Last Updated**: 2026-09-26
+**Version**: 9.28
+**Last Updated**: 2026-10-17
 **Purpose**: Consolidated development guidelines for GitHub Agentic Workflows
 
 This document consolidates specifications from the scratchpad directory into unified developer instructions. It provides architecture patterns, security guidelines, code organization rules, and testing practices.
@@ -2980,6 +2980,7 @@ These files are loaded automatically by compatible AI tools (e.g., GitHub Copilo
 ---
 
 **Document History**:
+- v9.28 (2026-10-17): Maintenance tone scan — re-ran the established strong-marketing scan (`powerful`/`seamless`/`effortless`/`blazing`/`cutting-edge`/`state-of-the-art`/`world-class`/`battle-tested`/`amazing`/`intuitive`/`magical`/`leverage`/`comprehensive`/`extensive`/`robust`/`sophisticated`/`elegant`/`deep dive`/`significantly`/`nice-to-have`/`best of both worlds`/`easy to`/`easily`/`simply`/`out-of-the-box`/`enhance(d)`/`great`/`revolutionary`/`game-changing`/`first-class`/`supercharge`/`deeply`) plus the v9.27-added subjective-adjective terms (`gracefully`, `minimalistic`) across all 67 `scratchpad/` spec files (`dev.md`, `README.md`, and the `smoke-test-22422877284.md` test artifact excluded, consistent with established precedent). 0 new tone issues: the only matches are the previously-accepted legitimate usage unchanged from v9.26 (`EnhanceError()`/`enhanceError()` identifiers in `errors.md`; "deeply nested" factual test-structure descriptions in `testing.md` and `mods/jsonschema-go.md`; the two `no magic` negations in `mdflow.md`/`mdflow-comparison.md`; `pr-checkout-logic-explained.md`'s neutral "differs significantly" comparison; `safe-outputs-specification.md`'s self-critical "significantly outdated" disclaimer) and the newly-checked `gracefully` occurrences across `gosec.md`, `artifact-naming-compatibility.md`, `github-mcp-access-control-specification.md`, `pr-checkout-logic-explained.md`, `go-type-patterns.md`, `agents/hierarchical-agents.md`, `breaking-cli-rules.md`, and `safe-outputs-specification.md`, all of which describe concrete, substantiated fallback/error-handling behavior (each followed by the specific degraded behavior), not unsubstantiated marketing. Cross-checked this file's Related Documentation links against the full `scratchpad/` file listing: all 67 non-excluded files remain linked, 0 new files since v9.27. The formatting scan's widespread bold-style pseudo-heading backlog (`**Field**:` used as structural labels across 50+ files) remains deferred as its own dedicated pass per the v9.27 note — no files were touched for this in this run. This sandbox's git history for `scratchpad/` shows a single synthetic commit (`d3915e4`, "fix: redact CLI proxy startup diagnostics before publication (#69517)") that touches the entire tree with no incremental prior state to diff against, consistent with every prior run's note; "new since last run" was determined by content inspection (grep + Related-Documentation cross-reference) rather than git log. Cache-memory at `/tmp/gh-aw/cache-memory/` contained no prior `consolidation/latest.json` this run (git-internal cache directory only, no consolidation metadata persisted by any earlier run). Coverage: 67 scratchpad spec files scanned (100%), 0 new files, 0 files touched.
 - v9.27 (2026-09-26): Ran the tone-analyzer's "great/easy/powerful/amazing/simple/seamless/intuitive/subjective-adjective" scan across 13 previously-clean `scratchpad/` spec files and fixed 34 tone issues missed by earlier marketing-keyword scans (unhedged "clear"/"helpful"/"complete"/"minimalistic"/"gracefully"/"full" claims, and "Key Insight"-style subjective section labels) across `repo-memory.md`, `guard-policies-specification.md`, `adding-new-engines.md`, `agent-sessions.md`, `labels.md`, `styles-guide.md`, `token-budget-guidelines.md`, `artifact-naming-compatibility.md`, `breaking-cli-rules.md`, `changesets.md`, and `schema-validation.md`; `metrics-glossary.md` and `capitalization.md` had none. Formatting scan flagged widespread bold-style pseudo-headings (`**Field**:`) used as structural labels in `repo-memory.md`, `metrics-glossary.md`, and `token-budget-guidelines.md` (100+ instances) — left as-is since converting them to `####` headings is a large mechanical change better handled as its own dedicated pass; flagged here for a future run. Coverage: 68 spec files (no new files; specs/ left out of scope per this run's "Specs Directory: scratchpad/" framing).
 - v9.26 (2026-09-12): Maintenance tone scan — re-ran the established strong-marketing scan (`powerful`/`seamless`/`effortless`/`blazing`/`cutting-edge`/`state-of-the-art`/`world-class`/`battle-tested`/`proven`/`amazing`/`intuitive`/`magical`/etc.) and the subjective-intensifier scan (`comprehensive`/`extensive`/`robust`/`flexible`/`smart`/`enhanced`/`enhance`/`deep`/`deeply`/`significantly`/`nice-to-have`/`best of both worlds`/`easy`/`easily`/`simply`/`sensible defaults`/`out-of-the-box`/etc.) across all 63 `scratchpad/` spec files (same exclusions as v9.25: `dev.md`, `README.md`, and the `smoke-test-22422877284.md` test artifact excluded; `specs/` left out of scope, consistent with this run's "Specs Directory: scratchpad/" framing). `scratchpad/` remains clean — 0 new tone issues. The scan's remaining matches are all previously-accepted legitimate usage, unchanged in wording from prior passes: `EnhanceError()`/`enhanceError()` function identifiers in `errors.md` (lines 130, 335, 344, 433, 652), not prose; "sensible defaults" describing `BaseEngine`/config defaults in `engine-architecture-review.md`, `engine-review-summary.md` (×2), `mdflow.md`, and `oh-my-code.md` — a concrete, accepted engineering term since v8.1–v8.4 fixed the surrounding phrasing but kept this term; "deeply nested" describing YAML/struct/error-chain test cases in `testing.md`, `mods/jsonschema-go.md`, and `errors.md` — factual technical description, not an intensifier; the two `no magic` negations in `mdflow.md` (line 229) and `mdflow-comparison.md` (line 553), unchanged since v9.18; `git-ai.md` line 156's "experimental until the sandbox contract is proven" — a normal engineering conditional, not a claim of an established track record; `pr-checkout-logic-explained.md` line 5's "differs significantly between `pull_request` and `pull_request_target` events" — a neutral factual comparison substantiated by the following explanation, not self-praise; and `safe-outputs-specification.md` line 10's "significantly outdated relative to the canonical specification" — a self-critical staleness disclaimer, not marketing. Cross-checked all 63 tracked spec file basenames against this file's Related Documentation and Document History text: every file is still referenced and no new `scratchpad/` files have appeared since v9.24's additions (`continual-experiments-mvp-plan.md`, `git-ai.md`, `mods/go-sdk.md`), so no new links were needed. This sandbox's git history remains a single synthetic snapshot commit (`b9caa39`, a large repo-wide sync unrelated to `scratchpad/` content), with no incremental history to diff against or verifiable PR to document this cycle, consistent with the v9.24/v9.22 notes. Cache-memory at `/tmp/gh-aw/cache-memory/` was not accessible from this sandbox this run (path restricted outside the repo working directory), consistent with every prior run's note; historical context was reconstructed from this Document History instead. Coverage: 63 scratchpad spec files scanned (100%), 0 new files, 0 files touched.
 - v9.25 (2026-09-05): Maintenance tone scan — ran a full-coverage scan (batched, ~10 files per pass) across all 63 `scratchpad/` spec files (excluding `dev.md`, `README.md`, and the `smoke-test-22422877284.md` test artifact per established precedent; `specs/` formal specification files with their own compliance fixtures were left out of scope for this pass, consistent with the "Specs Directory: scratchpad/" framing of this run) and fixed 7 formatting issues plus 3 tone issues across 5 files: `guard-policies-specification.md` (7 fixes: converted bold pseudo-headings under "GitHub Guard Policy Schema" and "Frontmatter Syntax" to `####` headings — "**Repos Scope:**" line 53, "**Minimal Example:**" line 72, "**With Repository Patterns:**" line 82, "**Public Repositories Only:**" line 95, "**Example - Public Repositories:**" line 163, "**Example - Specific Repositories:**" line 181, "**Implementation:**" line 204 — these read as section labels one level below their parent `### N.` headings, matching the v9.x precedent of converting bold-style headings to markdown syntax); `styles-guide.md` (3 fixes: added ` ```text ` language tags to the three ASCII color-swatch example blocks at lines 50, 75, and 97, previously untagged); `serena-tools-quick-reference.md` (1 fix: added ` ```text ` tag to the tool-call-breakdown bar chart at line 20); `template-syntax-sanitization.md` (1 fix: added ` ```text ` tag to the example warning-message block at line 63); `safe-output-handlers-refactoring.md` (1 fix: "All safe output handlers have been successfully refactored" → "...have been refactored" line 5 — "successfully" is redundant self-congratulation, the surrounding per-handler status table already substantiates completion); `engine-architecture-review.md` (1 fix: "✅ Simple, focused implementation" → "✅ Focused implementation" line 120 — "Simple" is an unsubstantiated subjective claim alongside the concrete "focused" descriptor); `actions.md` (1 fix: "Slightly slower execution (compilation overhead)" → "Increased execution time (compilation overhead)" line 380 — "Slightly" is a vague, unquantified qualifier for a stated trade-off). Left several scanner-flagged items unaddressed as false positives or low-value churn: `visual-regression-testing.md`'s "✅ **Update when**:" / "❌ **Don't update when**:" (lines 233, 239) are meaningful checkmark-prefixed list labels, not disguised headings — converting them would remove the ✅/❌ visual scanning aid the list depends on; long-list-to-prose suggestions across ~15 files (e.g., `testing.md` fuzz-test list, `template-injection-prevention.md` untrusted-input list, file listings in `actions.md`/`agent-sessions.md`) were not converted since these are legitimate reference/checklist content where prose would reduce scannability, not improve it; borderline word-choice flags in engineering rationale sections (e.g., "Better resource utilization" in `hierarchical-agents-quickstart.md`, "easier to maintain" in `validation-refactoring.md`, "Faster iteration" in `adding-new-engines.md`) were left as-is — these describe genuine engineering trade-offs rather than unsubstantiated marketing claims, and churning them file-by-file would not improve technical precision. No new files appeared in `scratchpad/` since v9.24's 76-file coverage baseline (73 previously tracked + 3 added last run), so no new Related Documentation links were needed this pass. Cache-memory at `/tmp/gh-aw/cache-memory/` was not accessible from this sandbox (path restricted outside the repo working directory, consistent with every prior run's note); historical context was reconstructed from this Document History instead. Coverage: 63 scratchpad spec files scanned (100%), 0 new files, 5 files touched.