Hacktoberfest 2026: le issue che i maintainer hanno segnato per ottobre, aperte e adatte ai principianti. Sfoglia le issue Hacktoberfest

Clarify agent query workflow across instructions, skill, CLI, and MCP

Aperta
  • Qualcosa non va?
#3,768 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
45/100
Tipo di issue
Documentazione
Chiarezza
Da chiarire
Stato di attività
Attiva
Stack tecnologico
python
Ambito
api, cli, developer-experience, documentation

Direzione di ricerca

Compare AGENTS.md, SKILL.md, references/query.md, LESSONS.md, the graphify query entry point, and the MCP query_graph tool schema. Trace how graphify-out/.vocab.txt is produced and used, then document the intended loading order, CLI/MCP boundaries, lesson handling, and query verification criteria. Done means the supported workflow and expected retrieval behavior are explicit and consistent.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

I’m using Graphify 0.9.65 with OpenCode and an existing graph. I cannot determine which query workflow an agent is supposed to follow because its instructions point in different directions:

  • The generated AGENTS.md guidance and the skill’s fast path say to run graphify query "<question>" directly.
  • The skill’s references/query.md says vocabulary expansion is required before traversal, then instructs the agent to query with the expanded terms. Loading SKILL.md lists this reference but does not load its contents. I tested the ordinary skill workflow with both Qwen 3.8 and Claude Sonnet 5; neither agent loaded the query reference on its own. I had to explicitly tell an agent to read it.
  • The reference regenerates graphify-out/.vocab.txt by reading all of graph.json as part of each query. The vocabulary is derived from node labels, so it should change only when the graph changes; selecting terms from it is the per-question step. On my ~670 MB graph, this makes every query repeat a large read and vocabulary build. I could find no CLI/MCP code that uses .vocab.txt or generates it during graph build/update. Is this file meant to be a persistent, graph-versioned artifact? If not, why must it be regenerated for every query, and how should agents tell whether an existing copy is current?
  • The reference says to read LESSONS.md at the start and begin with preferred sources, but does not explain how those sources affect query selection.
  • The reference describes CLI commands; the MCP server instead exposes query_graph and related tools, without vocabulary, reflect, or save-result tools. The skill does not say whether an agent should combine MCP calls with CLI steps or choose one interface.

This leaves an agent choosing between its AGENTS.md instructions, the main skill, a reference it does not reliably read, and the MCP tool schema. The website’s direct natural-language query examples add to the uncertainty. The query reference itself says the CLI has no stemming or synonyms; without its vocabulary procedure, ordinary natural-language wording that differs from graph labels can miss the relevant nodes. I do not yet know whether following the procedure consistently produces accurate results either. Could you explain what retrieval quality users should expect in both cases, and how to verify that a query started from the right nodes?

What is the intended flow for an agent answering a question against an existing graph? Please specify the order of skill/reference loading, vocabulary expansion, lessons, query, source checking, and feedback. In particular, where and how should lessons learned influence future queries? Are they intended to affect seed selection automatically, or must the agent apply them manually? Is the Graphify MCP server currently supported by this skill’s query workflow? If so, how should an MCP-based agent perform the CLI-only steps? If not, should the skill explicitly direct agents to use the CLI?

The separate OpenCode /graphify slash-command issue is already tracked in #2709.

Lingua principale
Python
Stelle
124k
Fork
11.9k
Nessuna PR unita negli ultimi 30g

Preparare l'ambiente

Questo progetto non fornisce container di sviluppo, Dockerfile né guida per i contributori, quindi l'ambiente è a tuo carico: parti dal suo README e consulta la nostra guida al primo contributo per i passaggi generali.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di Graphify-Labs/graphify

Tutte le issue di Graphify-Labs/graphify

Issue simili

Altre issue su Python

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.