Agent skill · software engineering · steipete
gitcrawl
GitHub archive: issue/PR search, sync freshness, duplicate clusters, current GitHub handoff, and Gitcrawl repo work.
Why this skill is useful
Adds specific commands and workflows for managing GitHub issues and PRs that the AI wouldn't reliably generate on its own.
What it needs
Requires gitcrawl installed locally. About 1k tokens when loaded. Last updated 2026-08-07. 385,400 stars on the source repository.
What this skill does
Gitcrawl Use local GitHub archives for discovery; verify current GitHub state before a maintainer decision or public mutation. Check freshness and configured database paths with gitcrawl doctor --json. For stale results, use a bounded --sync-if-stale 5m search or targeted gitcrawl sync owner/repo --numbers 123,456 --with pr-details. The latter hydrates local PR files, commits, checks, workflow runs, and review-thread resolution; it does not populate the separate GitHub CLI cache. Current GitHub State Use bare PATH gh and narrow native JSON fields: gitcrawl gh was removed. Its exit 2 is a migration notice, not an auth failure. Do not pass its --live/--cached flags or old field names to gh, restore an old Gitcrawl, bypass the existing Octopool shim, or change auth/PATH/config to repair that removed command. Octopool owns the shared gh cache; Gitcrawl owns the local archive. See the upstream migration guide when setting up or diagnosing that handoff. Archive Boundaries Archive state and similarity are candidate evidence, not proof of a duplicate or resolved issue. Report repository, issue/PR numbers, absolute date spans, cluster IDs, freshness, and known gaps. close-thread, close-cluster, and canonical-member choices are local overrides; they do not mutate GitHub. Public mutations use authorized gh workflows and checkout proof. For exact archive counts, Gitcrawl has no sql subcommand. Resolve .dbpath from doctor --json and use sqlite3 -readonly; do not modify database rows instead of using the maintainer commands. For Gitcrawl implementation work, verify the checkout remote is openclaw/gitcrawl and use its own instructions.
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills steipete/gitcrawl