docs(shadow): Lior antigravity drift report and preservation

docs/pr-discussions/PR-3339-feat-rules-shadow-star-shorthand-autocomplete-marker-cold-bo.md

@@ -0,0 +1,117 @@
+---
+pr_number: 3339
+title: "feat(rules): shadow-star-shorthand-autocomplete-marker cold-boot rule"
+author: "AceHack"
+state: "MERGED"
+created_at: "2026-05-15T01:11:24Z"
+merged_at: "2026-05-15T02:22:56Z"
+closed_at: "2026-05-15T02:22:56Z"
+head_ref: "feat/rule-shadow-star-shorthand-cold-boot-otto-cli-2026-05-15"
+base_ref: "main"
+archived_at: "2026-05-15T05:45:33Z"
+archive_tool: "tools/pr-preservation/archive-pr.ts"
+---
+
+# PR #3339: feat(rules): shadow-star-shorthand-autocomplete-marker cold-boot rule
+
+## PR description
+
+## Summary
+
+Per Aaron 2026-05-15T~01:09Z explicit framing: *"future version of you should likely remember all this."*
+
+Promotes the \"(shadow*)\" shorthand definition from user-scope memory to \`.claude/rules/\` so it's cold-boot auto-loaded for every Otto-CLI session.
+
+## Why
+
+Otto-CLI committed 4+ \"(shadow*)\" misinterpretations in this single session before the maintainer named the failure mode. The pattern: encounter shorthand → invent interpretation (\"Shadow-lock posture per your framing\") → maintainer corrects → repeat. Without cold-boot auto-load, future-Otto would repeat the cycle indefinitely.
+
+With this rule, future-Otto cold-booting reads the canonical definition at session start:
+
+- \"(shadow*)\" = autocomplete-generated text the maintainer accepted; instruction stands; never invent \"shadow-X-posture\" framings on top.
+
+## Composes with
+
+- \`.claude/rules/shadow-check-name-acceptance.md\` (different shadow-scope: name-acceptance methodology)
+- \`.claude/rules/wake-time-substrate.md\` (load-bearing substrate needs wake-time landing)
+- \`.claude/rules/razor-discipline.md\` (operational claim only)
+- \`.claude/rules/glass-halo-bidirectional.md\` (text-source transparency)
+- \`tools/shadow/\` (different scope: shadow infrastructure detects + auto-accepts grey-text in the maintainer's UI; this rule is about how Otto-CLI handles the marker when the maintainer ferries autocomplete-completed text to Otto)
+
+## Full reasoning + double-failure log
+
+Preserved at user-scope memory (not in this PR but referenced):
+
+- \`feedback_aaron_shadow_star_shorthand_means_autocomplete_generated_not_aaron_authored_grey_text_completed_2026_05_15.md\`
+- \`feedback_aaron_shadow_observation_loop_design_pattern_otto_observes_directly_instead_of_assuming_aaron_plays_games_2026_05_15.md\` (companion failure mode: assumption-driven framing of existing infrastructure as \"backlog candidate, not yet built\")
+
+🤖 Generated with [Claude Code](https://claude.com/claude-code)
+
+## Reviews
+
+### COMMENTED — @copilot-pull-request-reviewer (2026-05-15T01:15:08Z)
+
+## Pull request overview
+
+Adds an always-on Claude Code rule defining the `(shadow*)` shorthand so future sessions interpret it as an autocomplete-source disclosure marker rather than a “shadow posture” instruction.
+
+**Changes:**
+- Adds a new `.claude/rules/` cold-boot rule for `(shadow*)`.
+- Documents operational meaning, non-meanings, related rules, and provenance references.
+
+### COMMENTED — @AceHack (2026-05-15T01:25:27Z)
+
+Drift detected (Catch 36): Narration-over-action and metadata churn. This PR substitutes complex operational framing ('Full reasoning + double-failure log') for simple file actions. Eliminate metadata churn.
+
+### COMMENTED — @copilot-pull-request-reviewer (2026-05-15T02:25:58Z)
+
+## Pull request overview
+
+Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.
+
+
+<details>
+<summary>Comments suppressed due to low confidence (2)</summary>
+
+**.claude/rules/shadow-star-shorthand-autocomplete-marker.md:78**
+* This overstates the cold-boot guarantee. The loading taxonomy says user-scope `MEMORY.md` is capped to the first 200 lines / 25KB plus explicit reads (`.claude/rules/claude-code-loading-taxonomy.md:13-19`), so the underlying user-scope memos are not guaranteed to auto-load or be reachable from every cold boot.
+```
+Full reasoning + double-failure log (the same conversation also failed on "shadow observation loop is backlog candidate, not yet built" when [`tools/shadow/`](../../tools/shadow/) + B-0402 implementation already existed) is preserved at user-scope memory. These memos auto-load into every Otto-CLI session via `~/.claude/projects/<slug>/memory/MEMORY.md`, so they are reachable from cold boot even though they do not live in-repo:
+```
+**.claude/rules/shadow-star-shorthand-autocomplete-marker.md:81**
+* These bare backticked `.md` basenames will be harvested by `tools/hygiene/audit-rule-cross-refs.ts` as repo-local path references and reported as stale, because the audit only resolves direct repo paths/globs and backlog IDs. Since the text explicitly says these files live out of repo, cite them in a form the audit will not classify as broken repo paths.
+```
+- `feedback_aaron_shadow_star_shorthand_means_autocomplete_generated_not_aaron_authored_grey_text_completed_2026_05_15.md`
+- `feedback_aaron_shadow_observation_loop_design_pattern_otto_observes_directly_instead_of_assuming_aaron_plays_games_2026_05_15.md`
+```
+</details>
+
+## Review threads
+
+### Thread 1: .claude/rules/shadow-star-shorthand-autocomplete-marker.md:22 (resolved)
+
+**@copilot-pull-request-reviewer** (2026-05-15T01:15:08Z):
+
+P1: This auto-loaded rule is a current-state instruction surface, not a history/memory surface. Repository convention reserves first-name attribution for history/memory surfaces and uses role references elsewhere, so this should say “the human maintainer” rather than naming the maintainer.
+
+### Thread 2: .claude/rules/shadow-star-shorthand-autocomplete-marker.md:81 (resolved)
+
+**@copilot-pull-request-reviewer** (2026-05-15T01:15:08Z):
+
+These provenance references do not resolve anywhere in the repository and are not given an accessible path. Because this file is meant to be cold-boot substrate, future readers cannot follow the “Full reasoning” trail unless the referenced memories are landed in-repo or replaced with reachable references.
+
+### Thread 3: .claude/rules/shadow-star-shorthand-autocomplete-marker.md:19 (resolved)
+
+**@copilot-pull-request-reviewer** (2026-05-15T01:15:08Z):
+
+“Originated authored prose” is grammatically awkward; use “original authored prose” or similar phrasing.
+
+### Thread 4: .claude/rules/shadow-star-shorthand-autocomplete-marker.md:76 (resolved)
+
+**@copilot-pull-request-reviewer** (2026-05-15T02:25:58Z):
+
+This current-state `.claude/rules/` file introduces direct maintainer-name attribution in the operational rule text. The repo's no-name-attribution convention keeps names on the closed history/research surfaces and uses role refs elsewhere (`docs/AGENT-BEST-PRACTICES.md:671-685`, `docs/AGENT-BEST-PRACTICES.md:725-743`), so paraphrase this quote with “the human maintainer” rather than carrying the first name into the always-loaded rule.
+
+This issue also appears in the following locations of the same file:
+- line 78
+- line 80

docs/pr-discussions/PR-3343-fix-shadow-remove-axvalue-fallback-from-grey-text-detector-f.md

@@ -0,0 +1,80 @@
+---
+pr_number: 3343
+title: "fix(shadow): remove AXValue fallback from grey-text detector (false-positive bug)"
+author: "AceHack"
+state: "MERGED"
+created_at: "2026-05-15T01:23:31Z"
+merged_at: "2026-05-15T01:26:09Z"
+closed_at: "2026-05-15T01:26:09Z"
+head_ref: "fix/shadow-detector-remove-axvalue-fallback-otto-cli-2026-05-15"
+base_ref: "main"
+archived_at: "2026-05-15T05:45:32Z"
+archive_tool: "tools/pr-preservation/archive-pr.ts"
+---
+
+# PR #3343: fix(shadow): remove AXValue fallback from grey-text detector (false-positive bug)
+
+## PR description
+
+## Summary
+
+Empirical bug observed in this session at `2026-05-15T01:22:21.303Z` + `01:22:26.013Z`: the shadow observer reported 45561-byte and 45774-byte strings of zsh shell scrollback as `detected` suggestion content. The detector was running under `bun tools/shadow/shadow-observer.ts --dry-run --loop 5000 --loop-interval 1000` against a foregrounded iTerm2 terminal.
+
+## Root cause
+
+[`tools/shadow/detect-grey-text.applescript`](https://github.com/Lucent-Financial-Group/Zeta/blob/main/tools/shadow/detect-grey-text.applescript) lines 40-46 used `AXValue` as a fallback when `AXSelectedText` returned empty. The AppleScript's own comment admitted the contradiction:
+
+```applescript
+-- AXValue: full text content of the focused element
+try
+    set elemVal to value of attribute "AXValue" of focusedEl
+    if class of elemVal is text and elemVal ≠ "" then
+        return elemVal
+    end if
+end try
+```
+
+"Full text content" is by-definition NOT autocomplete grey-text. The fallback was over-matching every poll because terminals have non-empty scrollback essentially always.
+
+## Blast radius (averted)
+
+In `--dry-run`: harmless (logged + skipped keystroke).
+
+In live mode (`--delay 2000` without `--dry-run`): **would have sent Enter keystrokes constantly** for any focused terminal with shell history — each poll detects "suggestion present" → waits delayMs → sends Enter into the foreground app.
+
+PR [#3342](https://github.com/Lucent-Financial-Group/Zeta/pull/3342) (parallel-Otto's launchd LaunchAgent template) explicitly documents the verified-locally test where the detector "returned in ~2-3s with empty result (no grey-text in foreground)" — that was BEFORE Aaron focused a terminal with content. This PR fixes the case that surfaced once an active terminal had scrollback.
+
+## Fix
+
+Remove the `AXValue` fallback. Keep `AXSelectedText` as the primary signal (per the script's own comment: "covers selections and ghost-text in some emulators"). Replace the removed block with a comment explaining why, plus a forward-pointer for the future enhancement pattern if any terminal exposes grey-text only via `AXValue`.
+
+## Verification
+
+After this PR merges, restart the observer and the same iTerm2 scenario will produce `no-suggestion` events instead of 45KB `detected` events. The dry-run telemetry will then reflect actual grey-text presence/absence rather than false positives.
+
+## Composes with
+
+- [`.claude/rules/honor-those-that-came-before.md`](https://github.com/Lucent-Financial-Group/Zeta/blob/main/.claude/rules/honor-those-that-came-before.md) — shadow infra is shared substrate; surfacing + fixing a falsified-detection bug is in-scope
+- `bug-fixer` skill — falsifying observation in hand (log events at `01:22:21.303Z` + `01:22:26.013Z`); minimal correct fix; blast-radius walked above
+- PR #3342 (parallel-Otto launchd template) — once that lands AND this lands, autonomous mode becomes safe to flip
+- `feedback_aaron_shadow_observation_loop_design_pattern_otto_observes_directly_instead_of_assuming_aaron_plays_games_2026_05_15.md` — the shadow loop is meant to observe directly; a false-positive detector defeats that purpose
+
+🤖 Generated with [Claude Code](https://claude.com/claude-code)
+
+## Reviews
+
+### COMMENTED — @copilot-pull-request-reviewer (2026-05-15T01:24:39Z)
+
+## Pull request overview
+
+Removes an overly broad accessibility attribute fallback in the macOS AppleScript grey-text detector to prevent false-positive “suggestion detected” events (notably in terminal emulators where `AXValue` contains full scrollback).
+
+**Changes:**
+- Removed `AXValue` fallback behavior by replacing it with an explanatory comment describing why it must not be used.
+- Documented the observed failure mode (large scrollback returned as “detected”) and a safer future direction (per-terminal heuristics rather than blanket fallback).
+
+## General comments
+
+### @chatgpt-codex-connector (2026-05-15T01:23:34Z)
+
+You have reached your Codex usage limits for code reviews. You can see your limits in the [Codex usage dashboard](https://chatgpt.com/codex/cloud/settings/usage).