Round 44 batch 6a of 6: DV-2.0 provenance frontmatter — backfill script + 10 skills by AceHack · Pull Request #86 · Lucent-Financial-Group/Zeta

AceHack · 2026-04-22T03:53:54Z

Summary

Batch 6a of the 6-batch speculative-branch drain plan
(docs/research/speculative-branch-landing-plan-2026-04-22.md).
This is the skill-tune-up sub-batch of batch 6: lands
the Data-Vault-2.0 provenance-frontmatter backfill tooling
plus the first alphabetical batch of 10 skills as dry-run
validation.

What this lands

Script:

tools/skill-catalog/backfill_dv2_frontmatter.sh (207 lines)
— idempotent backfill that computes:
- record_source from git log --reverse first-land
  commit subject (regex [Rr]ound *([0-9]+) →
  "skill-creator, round N"; else fallback
  "git: <author> on <date>")
- load_datetime from first-land commit
- last_updated = today
- status defaults to "active" (honest default; a
  stub/dormant skill keeps the default until human
  review flips it)
- bp_rules_cited = [BP-11] (default floor for
  skills that handle audited surfaces)
- Usage: [--dry-run] <SKILL.md path>... | --all
- Re-running on a compliant file is a no-op.

Skills brought into compliance (10):

.claude/skills/activity-schema-expert/SKILL.md
.claude/skills/agent-experience-engineer/SKILL.md
.claude/skills/agent-qol/SKILL.md
.claude/skills/ai-evals-expert/SKILL.md
.claude/skills/ai-jailbreaker/SKILL.md
.claude/skills/ai-researcher/SKILL.md
.claude/skills/alerting-expert/SKILL.md
.claude/skills/algebra-owner/SKILL.md
.claude/skills/alignment-auditor/SKILL.md
.claude/skills/alignment-observability/SKILL.md

Compliance trajectory

After this commit: 12 of 216 SKILL.md files compliant
(the two that already had frontmatter plus this
alphabetical-batch-1 of 10). 204 remaining queued for
batches across future ticks — batches 6a+1, 6a+2, etc.
can re-run the same script on the next alphabetical
slice.

Drain-PR pre-check

Ran the memory/(user|feedback|project|reference)_|\baaron\b
grep on commit 5f6308f (cherry-pick of 48544ac) —
0 hits. Clean batch.

Composition

docs/DECISIONS/2026-04-21-data-vault-2-in-skill-catalog.md
— DV-2.0 adoption ADR (already landed upstream).
.claude/skills/skill-documentation-standard/SKILL.md
— standard that declares the five required
frontmatter fields.
BACKLOG row "DV-2.0 provenance rollout beyond the
skill catalog" — phase-1 deliverable logged in
commit a103f08 (already landed).

Test plan

Cherry-pick of 48544ac clean
Pre-check grep on commit 5f6308f — 0 hits
Script is shellcheck-clean at authoring time
(per original commit message)
CI green (markdownlint + shellcheck + editorconfig)
Re-running backfill_dv2_frontmatter.sh on any of
the 10 compliant files produces a no-op diff
Auto-merge resolves when CI clears

Phase-1 deliverable of the BACKLOG row "Data Vault 2.0 provenance as scope-universal indexing substrate — rollout beyond the skill catalog" (landed a103f08). Lands the mechanical cascade script and applies it to the first alphabetical batch of 10 skills as dry-run validation. Script: tools/skill-catalog/backfill_dv2_frontmatter.sh - Usage: [--dry-run] <SKILL.md path>... | --all - Idempotent (re-running on compliant file is a no-op) - Computes record_source from git-log --reverse first-land commit subject: regex [Rr]ound *([0-9]+) → "skill-creator, round N"; else fallback "git: <author> on <date>" - load_datetime from first-land commit; last_updated = today; status defaults to "active" (no inference beyond the honest default, a stub/dormant skill keeps the default until human review flips it); bp_rules_cited from grep BP-[0-9]+ → YAML inline list (`[]` if none) - Injects missing fields before the closing `---` fence using awk with ENVIRON (awk's -v flag refuses multi-line values) Batch 1 (10 skills, alphabetical a* prefix): activity-schema-expert (round 34, [BP-11]) agent-experience-engineer (round 34, [BP-01,03,07,08,11,16]) agent-qol (round 29, [BP-11]) ai-evals-expert (round 34, [BP-11]) ai-jailbreaker (round 34, [BP-11]) ai-researcher (round 34, []) alerting-expert (round 34, [BP-11]) algebra-owner (git: Aaron Stainback on 2026-04-18, []) alignment-auditor (round 37, [BP-10, BP-11]) alignment-observability (round 37, []) Catch-in-tick: initial script draft used `awk -v blob="$(printf ...)"` which corrupted activity-schema-expert/SKILL.md to 0 bytes on first application (awk -v does not accept multi-line values). Caught before any commit; reverted via git checkout and switched to ENVIRON. Re-ran, verified idempotency on second pass, then scaled to the 9-file batch. Compliance state after this commit: 12 of 216 SKILL.md files compliant (the two from prior ticks plus this batch of 10). 204 remaining queued for batches across future ticks. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

…atch6a

Copilot

Pull request overview

Adds a DV-2.0 provenance-frontmatter backfill script for .claude/skills/**/SKILL.md and uses it to backfill the required provenance fields on the first alphabetical batch of 10 skills, as part of the speculative-branch drain plan.

Changes:

Introduces tools/skill-catalog/backfill_dv2_frontmatter.sh to inject missing DV-2.0 frontmatter fields from git history (with --dry-run and --all modes).
Backfills DV-2.0 frontmatter fields (record_source, load_datetime, last_updated, status, bp_rules_cited) in 10 existing skill files.
Establishes initial dry-run validation coverage for the compliance rollout process.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
tools/skill-catalog/backfill_dv2_frontmatter.sh	New backfill script to compute/inject DV-2.0 provenance fields into SKILL.md frontmatter.
.claude/skills/activity-schema-expert/SKILL.md	Adds DV-2.0 provenance frontmatter fields.
.claude/skills/agent-experience-engineer/SKILL.md	Adds DV-2.0 provenance frontmatter fields.
.claude/skills/agent-qol/SKILL.md	Adds DV-2.0 provenance frontmatter fields.
.claude/skills/ai-evals-expert/SKILL.md	Adds DV-2.0 provenance frontmatter fields.
.claude/skills/ai-jailbreaker/SKILL.md	Adds DV-2.0 provenance frontmatter fields.
.claude/skills/ai-researcher/SKILL.md	Adds DV-2.0 provenance frontmatter fields.
.claude/skills/alerting-expert/SKILL.md	Adds DV-2.0 provenance frontmatter fields.
.claude/skills/algebra-owner/SKILL.md	Adds DV-2.0 provenance frontmatter fields (including a problematic `record_source`).
.claude/skills/alignment-auditor/SKILL.md	Adds DV-2.0 provenance frontmatter fields.
.claude/skills/alignment-observability/SKILL.md	Adds DV-2.0 provenance frontmatter fields.

Copilot · 2026-04-22T03:57:44Z

+  local file="$1" rules
+  rules=$(grep -oE 'BP-[0-9]+' "$file" 2>/dev/null | sort -u | paste -sd, - | sed 's/,/, /g')
+  if [[ -z "$rules" ]]; then
+    echo "[]"
+  else
+    echo "[${rules}]"


compute_bp_rules uses grep in a pipeline under set -euo pipefail. When a SKILL.md contains no BP-<NN> matches, grep exits 1, which makes the whole pipeline fail and can abort the script instead of producing []. Make this pipeline non-fatal on “no matches” (e.g., treat exit 1 as empty output) so skills without BP citations can be processed safely.

Suggested change

local file="$1" rules

rules=$(grep -oE 'BP-[0-9]+' "$file" 2>/dev/null | sort -u | paste -sd, - | sed 's/,/, /g')

if [[ -z "$rules" ]]; then

echo "[]"

else

echo "[${rules}]"

local file="$1" matches rules grep_status

if matches=$(grep -oE 'BP-[0-9]+' "$file" 2>/dev/null); then

:

else

grep_status=$?

if [[ $grep_status -eq 1 ]]; then

matches=""

else

return "$grep_status"

fi

fi

if [[ -n "$matches" ]]; then

rules=$(printf '%s\n' "$matches" | sort -u | paste -sd, - | sed 's/,/, /g')

echo "[${rules}]"

else

echo "[]"

Copilot · 2026-04-22T03:57:45Z

+# compute_record_source FILE -> "<author-heuristic>, round N"
+compute_record_source() {
+  local file="$1" subj
+  subj=$(git log --reverse --format='%s' -- "$file" 2>/dev/null | head -n 1)
+  if [[ "$subj" =~ [Rr]ound\ *([0-9]+) ]]; then
+    echo "skill-creator, round ${BASH_REMATCH[1]}"
+  else
+    # No round marker found — still honest: cite the author and date only.
+    local author_date
+    author_date=$(git log --reverse --format='%an on %ai' -- "$file" 2>/dev/null | head -n 1 | awk '{print $1" "$2" on "$4}')
+    echo "git: ${author_date:-unknown}"


compute_record_source’s fallback emits the git author’s personal name (e.g., git: <author> on <date>). This violates the repo’s “no name attribution in code/docs/skills” rule; record_source should use role-based identifiers (e.g., “human maintainer”, “architect”, “skill-creator/round-N”) rather than contributor names.

Suggested change

# compute_record_source FILE -> "<author-heuristic>, round N"

compute_record_source() {

local file="$1" subj

subj=$(git log --reverse --format='%s' -- "$file" 2>/dev/null | head -n 1)

if [[ "$subj" =~ [Rr]ound\ *([0-9]+) ]]; then

echo "skill-creator, round ${BASH_REMATCH[1]}"

else

# No round marker found — still honest: cite the author and date only.

local author_date

author_date=$(git log --reverse --format='%an on %ai' -- "$file" 2>/dev/null | head -n 1 | awk '{print $1" "$2" on "$4}')

echo "git: ${author_date:-unknown}"

# compute_record_source FILE -> role-based provenance string

compute_record_source() {

local file="$1" subj

subj=$(git log --reverse --format='%s' -- "$file" 2>/dev/null | head -n 1)

if [[ "$subj" =~ [Rr]ound\ *([0-9]+) ]]; then

echo "skill-creator, round ${BASH_REMATCH[1]}"

else

# No round marker found — keep provenance without personal-name attribution.

local first_date

first_date=$(git log --reverse --format='%ai' -- "$file" 2>/dev/null | head -n 1 | awk '{print $1}')

echo "git: human maintainer on ${first_date:-unknown}"

Copilot · 2026-04-22T03:57:45Z

+# compute_record_source FILE -> "<author-heuristic>, round N"
+compute_record_source() {
+  local file="$1" subj
+  subj=$(git log --reverse --format='%s' -- "$file" 2>/dev/null | head -n 1)
+  if [[ "$subj" =~ [Rr]ound\ *([0-9]+) ]]; then
+    echo "skill-creator, round ${BASH_REMATCH[1]}"


The record_source format produced for round-tagged commits ("skill-creator, round N") does not match the documented convention in skill-documentation-standard (skill-creator/round-N). If downstream tooling expects the standard shape, this mismatch will cause drift; consider emitting the standard tokenized form consistently.

Suggested change

# compute_record_source FILE -> "<author-heuristic>, round N"

compute_record_source() {

local file="$1" subj

subj=$(git log --reverse --format='%s' -- "$file" 2>/dev/null | head -n 1)

if [[ "$subj" =~ [Rr]ound\ *([0-9]+) ]]; then

echo "skill-creator, round ${BASH_REMATCH[1]}"

# compute_record_source FILE -> "skill-creator/round-N" or "git: <author> on <date>"

compute_record_source() {

local file="$1" subj

subj=$(git log --reverse --format='%s' -- "$file" 2>/dev/null | head -n 1)

if [[ "$subj" =~ [Rr]ound\ *([0-9]+) ]]; then

echo "skill-creator/round-${BASH_REMATCH[1]}"

Copilot · 2026-04-22T03:57:45Z

+  # Verify frontmatter is well-formed: exactly two `---` fences near the top.
+  local dash_count
+  dash_count=$(awk '/^---$/ { n++ } END { print n+0 }' "$file")
+  if [[ "$dash_count" -lt 2 ]]; then
+    echo "error: $file has no closing frontmatter fence" >&2
+    return 2
+  fi


Frontmatter validation counts every --- in the file, so a markdown horizontal rule later in the body can be mistaken for the closing frontmatter fence. This can cause the script to inject fields into the body and corrupt the file. Consider verifying that line 1 is --- and locating the closing fence within the initial frontmatter block (e.g., search only after the first line until the second fence).

Copilot · 2026-04-22T03:57:46Z

+
+RC=0
+for f in "${FILES[@]}"; do
+  process_one "$f" || RC=$?


process_one "$f" || RC=$? overwrites RC each time a later file fails, which can mask an earlier, more specific error (e.g., a malformed file returning 2 can be replaced by a later 1). Preserve the first non-zero exit code (or the max) so callers get a stable summary failure code when processing multiple files.

Suggested change

process_one "$f" || RC=$?

if ! process_one "$f"; then

status=$?

if [[ "$RC" -eq 0 ]]; then

RC=$status

fi

fi

Copilot · 2026-04-22T03:57:46Z

 ---
 name: algebra-owner
 description: Use this skill as the designated specialist reviewer for Zeta.Core's operator algebra — Z-sets, D/I/z⁻¹/H, retraction-native semantics, the chain rule, nested fixpoints, higher-order differentials. He carries deep advisory authority on the algebra's mathematical shape; final decisions require Architect buy-in or human sign-off (see docs/CONFLICT-RESOLUTION.md).
+record_source: "git: Aaron Stainback on 2026-04-18"


record_source includes a contributor’s personal name ("git: Aaron Stainback on 2026-04-18"), which violates the repo rule that code/docs/skills must avoid direct contributor names and use role references instead. Please replace this with a role-based identifier (and adjust the backfill script’s fallback so it can’t reintroduce names).

Suggested change

record_source: "git: Aaron Stainback on 2026-04-18"

record_source: "git: human maintainer on 2026-04-18"

Copilot · 2026-04-22T03:57:46Z

+record_source: "skill-creator, round 34"
+load_datetime: "2026-04-19"
+last_updated: "2026-04-21"
+status: active


This skill is explicitly described as dormant/gated off (“Currently gated OFF”), but the newly added frontmatter sets status: active. Per the skill documentation standard, dormant is the intended lifecycle state for gated-off skills (e.g., ai-jailbreaker until activation). Update status to reflect the documented gating.

Suggested change

status: active

status: dormant

10 named contributor archetypes (human + AI) for Zeta first-contact surfaces (issue templates, README, CONTRIBUTING.md, SECURITY.md, AGENTS.md). Companion to docs/README.md §Quick-start (document-reader audiences) and docs/EXPERT-REGISTRY.md (internal reviewers) — contributor personas answer "who just showed up wanting to contribute, and do we lose them in the first 90 seconds?" Scrubbed at pre-commit per drain-PR pre-check discipline: maintainer-name prose replaced with role-refs (BP-L284-L290); memory/ ref replaced with in-tree pointer to docs/BACKLOG.md P3 conversational- bootstrap row. Personas: typo-fixer / busy backend engineer / first-paper grad student / AI coding agent / systems engineer / security researcher / F# enthusiast / maintainer-external peer / factory-reuse adopter / returning contributor. Round-44 speculative-branch drain, batch 6b of 6 (6a skill tune-ups landed via #86; this is one of the small additive-only factory surfaces split out of batch 6 to land cleanly; 6c will carry the remaining anchor-doc subset). Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

Copilot AI review requested due to automatic review settings April 22, 2026 03:53

AceHack enabled auto-merge (squash) April 22, 2026 03:54

Merge remote-tracking branch 'origin/main' into land-skill-tune-ups-b…

c8e14d6

…atch6a

Copilot started reviewing on behalf of AceHack April 22, 2026 03:54 View session

AceHack merged commit 2c27692 into Lucent-Financial-Group:main Apr 22, 2026
11 checks passed

Copilot AI reviewed Apr 22, 2026

View reviewed changes

AceHack mentioned this pull request Apr 22, 2026

Round 44: CONTRIBUTOR-PERSONAS.md — batch 6b of 6 (additive-only) #87

Merged

3 tasks

AceHack mentioned this pull request Apr 22, 2026

Round 44: AGENTS.md — pointer to docs/AGENT-ISSUE-WORKFLOW.md (batch 6c of 6) #88

Open

3 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Round 44 batch 6a of 6: DV-2.0 provenance frontmatter — backfill script + 10 skills#86

Round 44 batch 6a of 6: DV-2.0 provenance frontmatter — backfill script + 10 skills#86
AceHack merged 2 commits intoLucent-Financial-Group:mainfrom
AceHack:land-skill-tune-ups-batch6a

AceHack commented Apr 22, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

-  local file="$1" rules
-  rules=$(grep -oE 'BP-[0-9]+' "$file" 2>/dev/null | sort -u | paste -sd, - | sed 's/,/, /g')
-  if [[ -z "$rules" ]]; then
-    echo "[]"
-  else
-    echo "[${rules}]"
+  local file="$1" matches rules grep_status
+  if matches=$(grep -oE 'BP-[0-9]+' "$file" 2>/dev/null); then
+    :
+  else
+    grep_status=$?
+    if [[ $grep_status -eq 1 ]]; then
+      matches=""
+    else
+      return "$grep_status"
+    fi
+  fi
+  if [[ -n "$matches" ]]; then
+    rules=$(printf '%s\n' "$matches" | sort -u | paste -sd, - | sed 's/,/, /g')
+    echo "[${rules}]"
+  else
+    echo "[]"

-  process_one "$f" || RC=$?
+  if ! process_one "$f"; then
+    status=$?
+    if [[ "$RC" -eq 0 ]]; then
+      RC=$status
+    fi
+  fi

	record_source: "git: Aaron Stainback on 2026-04-18"
	record_source: "git: human maintainer on 2026-04-18"

Conversation

AceHack commented Apr 22, 2026

Summary

What this lands

Compliance trajectory

Drain-PR pre-check

Composition

Test plan

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants