AGENT SKILL

workflows

When and how to use the repoindex plugin surface (MCP tools, agents, slash commands) for collection queries, release prep, activity summaries, and tag discipline. Use when users ask repoindex questions, mention their repo catalog, or want to know which repoindex tool fits their task.

License 许可
MIT
最近更新
2026-10-23
Stars
★ 0
Forks
0

01 前言 Front Matter

name
workflows
description
When and how to use the repoindex plugin surface (MCP tools, agents, slash commands) for collection queries, release prep, activity summaries, and tag discipline. Use when users ask repoindex questions, mention their repo catalog, or want to know which repoindex tool fits their task.

03 SKILL.md 正文

repoindex: which tool for which task

The plugin exposes three layers. Pick the smallest one that answers the question.

Layer Use when
MCP tool calls (run_sql, get_manifest) Single question, clear query, one answer.
Slash commands (/repo-week, /repo-status, /repo-audit, /repo-sprint, /repo-mirror) Named recurring task with a fixed shape.
Agents (repo-polish, repo-explorer) Multi-step reasoning, prose synthesis, or file writing.

Picking the layer

Use raw MCP when the user asks one question

"Show me my Python repos" is one run_sql. Do not invoke an agent. Do not reach for a slash command.

SELECT name, path FROM repos WHERE language = 'Python' ORDER BY name

Same for "how many repos do I have", "which repos are on PyPI", "what are my most starred repos". Direct SQL wins.

Use a slash command for recurring shaped queries

These have a known output format and the user invokes them by name:

  • /repo-week [window]: activity rollup via repoindex digest. Top churn, releases, totals.
  • /repo-status [name]: collection overview or single-repo detail card.
  • /repo-audit [filters]: read-only per-check summary of ops audit findings.
  • /repo-sprint: the triage entry point. "What needs attention / what's open?" (dirty, unpushed, active, release candidates, audit gaps, stale, unpublished).
  • /repo-mirror [flags]: dry-run-first wrapper around ops mirror for redundancy pushes.

Slash commands never write files on their own (mirror is the one exception, and only with --force-real).

Use an agent for multi-step or prose work

  • repo-polish: single-repo release prep. Audits, generates citation/license/codemeta, writes README improvements, syncs GitHub metadata. Writes files.
  • repo-explorer: open-ended analysis. Custom SQL across multiple tables with narrative synthesis.

Common patterns

Language-filtered audit

/repo-audit --language python --severity critical

Summary first. If the user wants to fix a specific repo, hand off to repo-polish.

Release prep flow

  1. /repo-status <name> to confirm state.
  2. repo-polish agent for the full audit-fix-improve cycle.
  3. /repo-audit --language <lang> after the release to confirm nothing regressed collection-wide.

Cross-platform status

SELECT r.name, p.registry, p.package_name, p.current_version, p.published, p.downloads_30d
FROM repos r JOIN publications p ON p.repo_id = r.id
WHERE p.published = 1
ORDER BY p.downloads_30d DESC NULLS LAST

Direct SQL. No agent needed. Join tags if filtering by classification.

Weekly standup for myself

/repo-week 7d

Or 30d for a monthly rollup.

Tag strategy

Tags live in the tags table. The source column says where a tag came from: user for tags assigned via mcp__repoindex__tag, everything else derived during refresh (repo state, forge topics, project keywords, registries). Query the current set rather than assuming it:

SELECT source, COUNT(*) FROM tags GROUP BY source

Naming conventions for user tags:

  • work/<area> for work-in-progress groupings: work/active, work/paused, work/release-queue.
  • Do not reuse the prefix: namespaces that derived tags occupy (lang:, topic:, keyword:, has:, published:; check with SELECT DISTINCT substr(tag, 1, instr(tag, ':')) FROM tags WHERE instr(tag, ':') > 0).
  • Project-domain tags: short nouns like ml, cli, dsl, paper, plugin.

Use mcp__repoindex__tag to add/remove user tags. Filter by tag with the --tag flag (accepted by copy, link, ops subcommands, and export; supports * wildcards, for example --tag 'work/*'), or from run_sql with the EXISTS (SELECT 1 FROM tags t WHERE t.repo_id = r.id AND t.tag = '...') pattern.

Anti-patterns to avoid

  • Do not invoke repo-explorer for a question that is one run_sql call.
  • Do not run /repo-sprint for single-repo questions. That is /repo-status or repo-polish.
  • Do not run repoindex ops commands without --dry-run on the first pass. /repo-mirror enforces this; other ops commands rely on the user choosing.
  • Do not bypass MCP. The CLI and the MCP server share a database, but MCP is the canonical read path inside Claude Code sessions.

Quick reference

Single question        → run_sql
Weekly rollup          → /repo-week
Glanceable status      → /repo-status
Read-only audit        → /repo-audit
Triage / kickoff       → /repo-sprint
Redundancy push        → /repo-mirror
Release prep (writes)  → repo-polish agent
Ad-hoc deep analysis   → repo-explorer agent

04 同仓库更多 Skills

Abdulqudduz
Abdulqudduz/claude-anvil

worldsmith-methodology

Use when working in a project with a worldbuilding documentation ecosystem (docs/ or lore/ directory). Provides the documentation-first methodology: canonical hierarchy, doc roles (timeline, lore, characters, systems), propagation discipline, and the dual canonical and exploratory workflow.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

prose-craft

Use when writing fiction, editing fiction prose, reviewing chapters, or drafting scenes. Covers prose craft, dialogue mechanics, scene structure, and common failure modes in AI-assisted fiction. Trigger phrases: "write fiction", "edit chapter", "prose review", "scene draft", "dialogue check", "fiction craft", "edit this scene", "review my prose".

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

vista

This skill should be used when the user asks about research directions, open problems, future work, or follow-up research from the academic literature. Trigger phrases include "open problems in X", "what's next for Y", "future work for paper Z", "research directions on T", "salient follow-up research", "what should I work on next in", "find me follow-up research unrelated to my prior work", "broad survey of W", "neglected directions in V". Routes the request to the right Vista MCP tool, reads back the structured paper sections, and synthesizes research directions in the conversation.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

synthesize

Make a research-agent run conclude and write synthesis.md (outcome, key findings, failed approaches, open questions, recommendations). Signals a live researcher to wrap up, or launches one in synthesis mode when none is running. Use when results are good enough or the agent is stalling.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

status

Show the current state of a research-agent run from .research/: liveness, cycles and budget, sub-problem and hypothesis statuses, eval trend, current focus, and branches. Read-only.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

resume

Resume an interrupted research-agent run from .research/. Re-launches the researcher to reload its structured state and continue from the current focus, optionally on a named branch or with a new budget. Use after a session restart, crash, or explicit pause.

★ 0