…y correction (#804) (#805)

* chore(loop-tick-history): tick 05:50Z — drain (3 PRs) + Aaron mid-tick durable-retry correction (#804)

3 CLEAN PRs squash-merged (#801/#802/#803). Aaron's /btw aside
answered (EAT document location). Mid-tick correction caught the
elan-toolchain-502 rerun anti-pattern from earlier this session;
durable fix landed in PR #804 (memory file + linux.sh + elan.sh
migrated to curl_fetch). First non-tick-history substrate this
session.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* chore(loop-tick-history): tick 0550Z — Copilot P2 fixes (memory/ prefix + explicit backlog paths)

PR #805 review thread from copilot-pull-request-reviewer:
1. Cross-reference style: memory file path needs `memory/`
   prefix (consistent with other shards).
2. "pending task #307" reads ambiguously (looks like a GH PR
   number). Replace with explicit backlog row paths
   (B-0062 + B-0074 full paths).

Both fixes applied; row content unchanged.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

 pending CI (#807)

Two CLEAN PRs merged. PR #804 (durable-retry fix) is now on main —
future external-dep download flakes absorbed via curl_fetch retry.
PR #806 (multi-AI feedback absorb bundle) waiting on CI — real-
dependency wait, not manufactured patience.

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

…+ tick 0558Z shard

The human maintainer forwarded a multi-AI synthesis packet during
autonomous-loop tick 05:58Z:
  - Deepseek's reassessment + 5-point pushback after correcting
    earlier-incorrect search results
  - Amara's filter-to-actionables (6 bounded items, with
    "rerun is incident recovery; retry/cache is substrate
    improvement" elevated as the load-bearing line)
  - reference to an older Gemini log on tele+port+leap operational
    resonance (already canonical at
    memory/feedback_operational_resonance_*.md; not re-absorbed)

Verbatim absorb (per the channel-verbatim-preservation rule)
landed at:
  docs/research/multi-ai-feedback-2026-04-29-deepseek-amara-on-loop-state.md
with §33 archive header (Scope / Attribution / Operational status /
Non-fusion disclaimer).

Four small P3 backlog rows file the bounded actionables (per
the maintainer's existing narrowing on task #309 — research-
grade only, no broad new substrate PRs):

  B-0098 — tick-ordinal-continuity lint (or remove ordinals
           from shards entirely; computed > narrated)
  B-0099 — PR-count claims as derived metrics, not narrated prose
  B-0100 — pure-wait tick backpressure / quiescence rule
  B-0101 — small 5-bucket reviewer-artifact classification table

The 5th actionable (external-dep retry/cache) is already
addressed by PR #804 (durable-retry fix landed alongside the
"rerun-is-recovery / retry-is-substrate-improvement" rule).
The 6th actionable (evidence-claim language tightening) is
operational discipline, captured in the tick shard observation
column.

Tick shard at docs/hygiene-history/ticks/2026/04/29/0558Z.md
captures the work-stream summary.

Pattern: ONE consolidated PR for the absorb bundle (research
note + 4 backlog rows + tick shard) rather than 6 separate PRs,
honoring the maintainer's "don't open a bunch of new PRs"
narrowing while still preserving the verbatim record durably.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

…age (B-0101 taxonomy applied) + rebase (#810)

(1) PR #808 squash-merged.
(2) #806's 5 unresolved threads classified per B-0101 taxonomy:
    - 1 DISPLAY_ARTIFACT (Copilot's "|| " excerpt was hallucinated;
      actual file has correct table)
    - 3 REVIEWER_SNAPSHOT_LAG (memory file IS on main post-#804 merge)
    - 1 REAL_DEFECT already fixed in prior tick
    All 5 resolved with explanatory comment.
(3) #806 branch 4 commits behind main → rebased + force-pushed;
    CI recomputing.

First operational use of the B-0101 taxonomy — one tick after
filing it. The pattern: file the rule, validate it on the next
class instance.

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

…ck 0558Z (re-open of #806) (#811)

* absorb: multi-AI feedback packet (Deepseek + Amara) + 4 backlog rows + tick 0558Z shard

The human maintainer forwarded a multi-AI synthesis packet during
autonomous-loop tick 05:58Z:
  - Deepseek's reassessment + 5-point pushback after correcting
    earlier-incorrect search results
  - Amara's filter-to-actionables (6 bounded items, with
    "rerun is incident recovery; retry/cache is substrate
    improvement" elevated as the load-bearing line)
  - reference to an older Gemini log on tele+port+leap operational
    resonance (already canonical at
    memory/feedback_operational_resonance_*.md; not re-absorbed)

Verbatim absorb (per the channel-verbatim-preservation rule)
landed at:
  docs/research/multi-ai-feedback-2026-04-29-deepseek-amara-on-loop-state.md
with §33 archive header (Scope / Attribution / Operational status /
Non-fusion disclaimer).

Four small P3 backlog rows file the bounded actionables (per
the maintainer's existing narrowing on task #309 — research-
grade only, no broad new substrate PRs):

  B-0098 — tick-ordinal-continuity lint (or remove ordinals
           from shards entirely; computed > narrated)
  B-0099 — PR-count claims as derived metrics, not narrated prose
  B-0100 — pure-wait tick backpressure / quiescence rule
  B-0101 — small 5-bucket reviewer-artifact classification table

The 5th actionable (external-dep retry/cache) is already
addressed by PR #804 (durable-retry fix landed alongside the
"rerun-is-recovery / retry-is-substrate-improvement" rule).
The 6th actionable (evidence-claim language tightening) is
operational discipline, captured in the tick shard observation
column.

Tick shard at docs/hygiene-history/ticks/2026/04/29/0558Z.md
captures the work-stream summary.

Pattern: ONE consolidated PR for the absorb bundle (research
note + 4 backlog rows + tick shard) rather than 6 separate PRs,
honoring the maintainer's "don't open a bunch of new PRs"
narrowing while still preserving the verbatim record durably.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* fix(B-0101): markdownlint MD022 — collapse split heading to single line

PR #806 markdownlint failure: heading was wrapped across two
`##` lines, which markdownlint reads as two separate headings
without blanks between them (MD022/blanks-around-headings).
Collapsed to single line.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* fix(B-0098/B-0099): Copilot P1 — portable grep boundary + explicit gh-login placeholder

Two real defects on PR #811:

B-0098 — \b is non-portable in POSIX ERE (grep -E). On GNU/BSD
  grep \b is treated as backspace/undefined. Replaced with -w
  (whole-word match, supported on both GNU and BSD grep) and
  added a comment documenting why.

B-0099 — `@me` reads ambiguously in pseudocode (looks like a
  literal token even though it IS valid GitHub search syntax
  for the authenticated user). Replaced with explicit
  `<gh-login>` placeholder + a clarifying note that `@me` also
  works.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* fix(B-0098/B-0099): convergent reviewer corrections (round 3)

Three external reviewers (Amara, Claude.ai, Deepseek) flagged
two precision issues on PR #811:

B-0098 — `grep -w` is GNU/BSD-common but not strictly POSIX-
  portable. Replace single-claim wording with two viable
  patterns: (a) `grep -woE` (GNU/BSD-common) and (b) strict
  portable explicit-boundary pattern. Implementing contributor
  picks based on portability priority.

B-0099 — `author:@me` inside `--search` reads ambiguously and
  is not the documented CLI shape. Replace with `gh pr list
  --author` CLI flag, with both `<your-gh-login>` (explicit,
  preferred for cold readability) and `@me` (valid CLI
  shorthand) shown.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* fix(B-0101): split SNAPSHOT_MISMATCH into backward-stale + forward-dependent (Amara round-3+round-4)

Amara's correction: REVIEWER_SNAPSHOT_LAG was too broad —
covered both temporal directions. Split into:

  SNAPSHOT_MISMATCH (parent)
    ├─ BACKWARD_STALE_SNAPSHOT — reviewer behind reality
    └─ FORWARD_CROSS_PR_REFERENCE — PR references sibling
         work not yet on base; valid only IF merge order
         is enforced

Same family, different remedies. Backward = verify-and-
resolve. Forward = encode dependency + don't resolve as
"valid post-merge" unless mechanically enforced.

Distilled rule (Amara): "A forward reference is not wrong
if the dependency is enforced. A forward reference is wrong
if the dependency is only hoped."

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* fix(B-0098): Codex P1 — strict-POSIX example must use only POSIX features

Codex correctly flagged that my "strict POSIX-portable"
example used bash-only features:
- [[ ]] (bash test, not POSIX)
- [[ a == *b* ]] (bash glob match, not POSIX)
- $(ls -1 ... | sort) for iteration

Replaced with strict POSIX:
- direct glob iteration `for file in pattern; do`
- `case ... in pattern) ;; esac` for glob match
- `printf` instead of `warn` (warn is shell function, not POSIX)
- redirect to stderr (>&2)

Option (a) keeps bashisms since it's labeled "GNU/BSD-common"
(works on every 4-shell target including bash + zsh on
realistic systems).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>