refresh-index

name	refresh-index
description	Re-ingest every already-enabled repo in data/skill-index-resources.json: sync the working tree, run preindex, rebuild the catalog for verification, summarize updated/unchanged/failed/skipped, then gate commit + PR behind explicit confirmation. Use when asked to refresh the index, re-ingest indexed repos, or batch-maintain already-indexed skill sources. Don't use for adding new repos (use skill-index-updater), improving a single skill (use skill-auto-improver), or installing/updating skills on the local machine (use asm install or asm update).
license	MIT
compatibility	Claude Code
allowed-tools	Bash Read Write Edit Grep Glob
effort	high
metadata	{"version":"1.0.0","author":"luongnv89"}

Refresh Index

You are refreshing every enabled repository in the ASM curated skill index. The goal is to re-ingest each source so data/skill-index/{owner}_{repo}.json reflects the latest upstream state, classify each repo as updated / unchanged / failed / skipped, verify the catalog still rebuilds cleanly, and open a single PR that ships only the data changes — with explicit user confirmation before any commit or push.

This is the inverse of skill-index-updater. That skill adds new repos from URLs the user supplies. This skill refreshes the repos that are already enabled in data/skill-index-resources.json.

When to Use

• User asks to "refresh the index", "update the indexed skills", "sync the catalog", "re-ingest all repos", or "batch-maintain the skill index"
• A scheduled refresh is due and the user wants to bring data/skill-index/*.json up to date with upstream changes
• A new ASM release is approaching and the catalog should reflect the latest upstream skill metadata

Do not trigger for: adding a new repository to the index (use skill-index-updater), authoring or improving a single skill (skill-creator, skill-auto-improver), opening an upstream PR to a third-party skill (skill-upstream-pr), or installing/updating skills on the local machine (asm install, asm update).

Repo Sync Before Edits (mandatory)

Before re-ingesting anything, pull the latest remote branch:

branch="$(git rev-parse --abbrev-ref HEAD)"
dirty=0
if [ -n "$(git status --porcelain)" ]; then
  git stash push -u -m "pre-refresh-index: ${branch}"
  dirty=1
fi
git fetch origin
git pull --rebase origin "$branch"
if [ "$dirty" -eq 1 ]; then
  git stash pop || {
    echo "✗ Stash pop failed — recover with: git stash list && git stash show -p stash@{0}"
    exit 1
  }
fi

If origin is missing or rebase conflicts occur, stop and ask the user before continuing. Never silently overwrite local edits to data/skill-index/ or data/skill-index-resources.json.

Prerequisites

Verify each before any ingest. Stop and tell the user if any fails.

• bun on PATH (command -v bun) — required by bun run preindex
• asm on PATH (command -v asm) — invoked transitively by the ingester for asm eval
• gh on PATH and authenticated (gh auth status) — required for PR creation
• git on PATH and inside the ASM repo working tree (git rev-parse --show-toplevel)
• Network access to github.com (each ingest clones the upstream repo)

Pipeline

Follow these steps in order. Each step has a verification check — do not proceed if the check fails.

Step 1: Enumerate the index

Read data/skill-index-resources.json and split entries into two lists:

ROOT="$(git rev-parse --show-toplevel)"
RES="$ROOT/data/skill-index-resources.json"

Use jq (or your equivalent JSON reader) to build two lists, keyed on the source string (github:owner/repo) because that is the identifier preindex echoes in its log lines:

• enabled[] — every repo with "enabled": true. These will be refreshed.
• disabled[] — every repo with "enabled": false. These land in the skipped bucket up front with reason "disabled in skill-index-resources.json".

jq -r '.repos[] | select(.enabled == true)  | .source' "$RES"   # e.g. github:anthropics/skills
jq -r '.repos[] | select(.enabled == false) | .source' "$RES"

Also derive the on-disk file key ({owner}_{repo}) for each enabled repo — this is what data/skill-index/{owner}_{repo}.json is named:

jq -r '.repos[] | select(.enabled == true) | "\(.source)\t\(.owner)_\(.repo)"' "$RES"

Verification: both lists are non-empty (the index always contains at least one enabled repo and one disabled self-reference). Also validate the existing per-repo JSON is parseable before re-ingesting — a pre-existing corruption otherwise gets masked by the refresh:

jq empty data/skill-index/*.json   # exits non-zero if any file is invalid

If enabled[] is empty or pre-validation fails, stop — there is nothing safe to refresh.

Step 2: Snapshot pre-run skill counts

Before re-ingesting, record the current skillCount for each enabled repo. This lets Step 7 report skill-count deltas.

mkdir -p /tmp/refresh-index
SNAPSHOT="/tmp/refresh-index/pre-snapshot.json"
echo "{}" > "$SNAPSHOT"
for entry in $(jq -r '.repos[] | select(.enabled == true) | "\(.owner)_\(.repo)"' "$RES"); do
  file="$ROOT/data/skill-index/${entry}.json"
  if [ -f "$file" ]; then
    count=$(jq -r '.skillCount // 0' "$file")
    jq --arg k "$entry" --argjson v "$count" '. + {($k): $v}' "$SNAPSHOT" > "$SNAPSHOT.tmp" && mv "$SNAPSHOT.tmp" "$SNAPSHOT"
  fi
done

If a repo is enabled but has no existing index file, treat the pre-count as 0. The repo will then show up in updated with a positive delta in Step 7.

Step 3: Run bun run preindex

Re-ingest every enabled repo via the project's preindex script. Capture both stdout and the exit code — preindex exits 1 if any repo fails, but do not abort the run. We still want partial results so the summary can show what succeeded.

LOG="/tmp/refresh-index/preindex.log"
cd "$ROOT"
set +e
bun run preindex 2>&1 | tee "$LOG"
PREINDEX_EXIT=$?
set -e

Verification: the log file exists and contains one line per enabled repo formatted as either {source} ... {N} skills (success) or {source} ... FAILED: {error} (failure), where {source} matches the github:owner/repo strings from Step 1. If the log is empty or no line matches, stop and surface the error.

Step 4: Classify each repo

Build per-repo status from three signals — the preindex log, git diff on data/skill-index/, and the disabled[] list from Step 1.

DIFF_FILES=$(git diff --name-only -- data/skill-index/ | sort -u)

For each enabled repo, match the preindex log line by its {source} string (github:owner/repo) and look up the on-disk file by {owner}_{repo}:

Signal in preindex.log (per {source} line)	data/skill-index/{owner}_{repo}.json in git diff	Bucket
{source} ... N skills	yes	updated
{source} ... N skills	no	unchanged
{source} ... FAILED: ...	(either)	failed (capture error message)
(no line for this {source})	(either)	failed (capture as "no output from preindex")

For each disabled[] repo: skipped with the disabled reason.

Capture each repo's post-run skillCount by reading the (possibly updated) data/skill-index/{owner}_{repo}.json. For failed repos, post-count = pre-count (no change on disk).

Step 5: Rebuild the website catalog (verification only)

Run the catalog build to confirm the refreshed index is structurally valid. website/catalog.json is gitignored — never stage it. A failure here means the data files are internally inconsistent and the PR must not land.

cd "$ROOT"
bun scripts/build-catalog.ts

Verification: the script exits 0 and website/catalog.json is valid JSON (jq empty website/catalog.json). If it fails, stop the pipeline — investigate the data issue before committing anything.

Step 6: Detect unexpected diff scope

Confirm the only files that changed are the ones we expect — per-repo data under data/skill-index/ and, only if the user explicitly bumped the top-level updatedAt, data/skill-index-resources.json:

UNEXPECTED=$(git diff --name-only \
  | grep -v -E '^data/skill-index/' \
  | grep -v -E '^data/skill-index-resources\.json$' \
  || true)
if [ -n "$UNEXPECTED" ]; then
  echo "⚠ Unexpected files in diff:"
  printf '%s\n' "$UNEXPECTED"
  # stop and surface to user
fi

Note: bun run preindex does not modify data/skill-index-resources.json — it only writes per-repo files under data/skill-index/. The resources file should only appear in the diff if the user is also refreshing the top-level updatedAt timestamp (optional — leave it alone unless asked).

If unexpected files appear (e.g. someone left edits in src/ or skills/), stop and tell the user. Do not commit a mixed change.

Step 7: Print the four-bucket summary

Render a markdown table grouped by bucket, with skill-count deltas. This is what the user reads to decide whether to confirm the PR.

## Refresh summary — N repos processed

### ✓ Updated (X)
| Repo | Before | After | Δ |
|------|--------|-------|---|
| anthropics/skills | 14 | 15 | +1 |
| obra/superpowers  | 22 | 22 |  0 |

### · Unchanged (Y)
| Repo | Skills |
|------|--------|
| owner1/repo1 | 7 |

### ✗ Failed (Z)
| Repo | Error |
|------|-------|
| owner2/repo2 | clone failed: 404 Not Found |

### ○ Skipped (W)
| Repo | Reason |
|------|--------|
| luongnv89/asm | disabled in skill-index-resources.json |

If X + Y + Z + W does not equal len(enabled) + len(disabled), the classification is inconsistent — stop and re-check Step 4 before moving on.

Step 8: Confirmation gate, commit, and PR

Do not proceed without explicit user confirmation. Print the diff stat and ask:

git diff --stat -- data/skill-index/ data/skill-index-resources.json
echo
echo "Ready to commit the files above and open a PR."
echo "Type 'yes' to continue, anything else to abort."

On yes (and only yes), stage only the index data files — never website/catalog.json, never anything outside data/skill-index/ — commit with the conventional-commit message below, push, and open the PR:

git add data/skill-index/
# Only add the resources file if it was intentionally modified (e.g., updatedAt bump):
if git diff --name-only | grep -q '^data/skill-index-resources\.json$'; then
  git add data/skill-index-resources.json
fi

git commit -m "$(cat <<'EOF'
chore(index): refresh indexed skill sources

Re-ingested all enabled repos in data/skill-index-resources.json.

Updated: <X> repo(s)
Unchanged: <Y> repo(s)
Failed: <Z> repo(s)
Skipped: <W> repo(s)
EOF
)"

branch="$(git rev-parse --abbrev-ref HEAD)"
git push -u origin "$branch"

gh pr create --title "chore(index): refresh indexed skill sources" --body "$(cat <<'EOF'
## Summary
Re-ingested all enabled repos in `data/skill-index-resources.json` to bring the catalog up to date with upstream.

## Results
- **Updated:** <X> repo(s)
- **Unchanged:** <Y> repo(s)
- **Failed:** <Z> repo(s) (see body for details)
- **Skipped:** <W> repo(s) (disabled in resources file)

### Updated repos
| Repo | Before | After | Δ |
|------|--------|-------|---|
| ... | ... | ... | ... |

### Failed repos
| Repo | Error |
|------|-------|
| ... | ... |

## Test Plan
- [ ] `data/skill-index/*.json` files are valid JSON
- [ ] `bun scripts/build-catalog.ts` rebuilds `website/catalog.json` without errors
- [ ] No files outside `data/skill-index/` and `data/skill-index-resources.json` are staged
- [ ] CI passes
EOF
)"

Fill the <X> / <Y> / <Z> / <W> placeholders and the per-bucket tables with the actual numbers from Step 7 before running gh pr create.

Verification: gh pr view --json url returns the new PR URL. Print it back to the user.

Expected Output

On a successful run, the user should see (and the agent should verify) all of the following:

1. Repo synced — branch is up to date with origin, any local edits were stashed and restored cleanly.
2. bun run preindex completed — exit code captured; per-repo lines visible in the log.
3. All four buckets populated — every enabled repo lands in exactly one of updated / unchanged / failed, and every disabled repo lands in skipped. Totals add up.
4. bun scripts/build-catalog.ts succeeded — website/catalog.json rebuilt and is valid JSON. Not staged.
5. Diff scope contained — only data/skill-index/*.json (and optionally data/skill-index-resources.json if explicitly refreshed) appear in git diff.
6. User confirmed — explicit yes recorded before commit + push.
7. PR opened — conventional-commit title (chore(index): refresh indexed skill sources), body filled from Step 8 template, URL returned to user.

If any of items 1–5 fails, do not proceed to Steps 6–7.

Edge Cases

Inputs the skill must handle without crashing, in order of likelihood:

• No enabled repos (enabled[] is empty): stop with an explanation — there is nothing to refresh.
• Existing local edits to data/skill-index/: the mandatory pre-edit stash (Repo Sync Before Edits) captures them; if the post-run pop conflicts, the recovery hint in the stash block tells the user how to restore.
• A single upstream repo is unreachable (404, network blip): preindex marks it FAILED and continues; the failed repo lands in the failed bucket. The PR still ships for the other repos.
• All upstream repos fail (no network, GitHub outage): every enabled repo lands in failed; the diff is empty; stop before Step 8 because there is nothing to commit.
• A repo's ingest produces zero skillCount (upstream removed every SKILL.md): treat it as updated with a negative delta — the change is real and worth shipping.
• bun run preindex is missing or errors before producing any log lines: stop in Step 3 — this is environmental, not a per-repo failure. Prompt the user to run bun install and retry.
• bun scripts/build-catalog.ts fails after a successful preindex: stop in Step 5 — the index is internally inconsistent. Investigate manually before committing anything.
• website/catalog.json accidentally appears in git status: it is gitignored; if .gitignore was broken, fix it before staging. Never git add website/catalog.json.
• Unexpected files in the diff (someone left WIP edits in src/ or skills/): stop in Step 6 and tell the user — do not commit a mixed change.
• User declines the confirmation in Step 8: stop cleanly. Leave the refreshed data/skill-index/*.json files in the working tree so the user can inspect or re-confirm later. Do not git checkout -- anything.
• gh CLI not authenticated: prompt the user to run gh auth login before retrying Step 8.

Error Handling

• Stash pop conflict after sync: surface the recovery hint, stop the pipeline. The user resolves manually.
• preindex exits 1 but produced a partial log: continue to Step 4 — partial results are still useful.
• build-catalog exits non-zero: stop. A broken catalog is a structural data issue, not a transient error.
• git diff shows unexpected files: stop. Ask the user to either revert unrelated edits or run this skill on a clean branch.
• gh pr create fails (auth, network, missing remote): print the staged commit SHA so the user can push and open the PR manually.

Cleanup

After the PR is opened (or the pipeline aborts), remove temporary artifacts:

rm -rf /tmp/refresh-index

Leave the working tree as the user left it. Do not git checkout anything they did not stage.