hypaware-query skill: Codex tool failures read as success; explain exit evidence and scoped duplicate counting

Aperta
  • Qualcosa non va?
#2,796 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
3/5
Tempo stimato
1-2 giorni
Idoneità per principianti
68/100
Tipo di issue
Documentazione
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
javascript
Ambito
documentation

Direzione di ricerca

Read hypaware-query/ai-messages.md and find its three host copies; inspect how the current guidance describes Codex tool results and repeated results. Update all three copies identically to explain recognized exit evidence, unknown outcomes, scoped duplicate counting, distinct attempts, unmatched rows, and the denominator. Then run the relevant skill test or add a fixture covering the five listed cases; done when the guidance and coverage match the issue's criteria.

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

Descrizione

Problem

hypaware-query/ai-messages.md does not explain how to count Codex tool failures, and the obvious query gets it wrong. Codex tool results carry their exit status in the output text, not in a structured error flag:

  • toolResultBlockFromPayload keeps the output text and call id but derives no is_error.
  • The shared message projector treats a tool_result with no is_error as success.

A synthetic Codex result whose output is JSON with exit_code: 1 projects with that JSON in content_text, its tool_call_id, and status.tool_status: success. So a query that counts tool_status errors misses recorded command failures. (Found while reviewing #2794; reviewer probe in review-r1-evidence/side-probes.log of that review.)

The same guide also says nothing about repeated copies of one tool result. The tempting fix, DISTINCT tool_call_id, is wrong: the projector scopes tool identities by session and conversation, and ids can repeat across threads and gateways.

Expected

ai-messages.md (all three host copies) explains:

  • Which Codex output shapes carry exit evidence, and that missing or unrecognized exit evidence is an unknown outcome, not success. Do not assume every tool has an exit_code field.
  • How to count a verified repeated copy of one tool result once, scoped by provider/gateway, session and conversation, keeping distinct attempts and unmatched rows, and stating the denominator. Never a global dedup by tool_call_id.

Done when

  • The guidance is in the three ai-messages.md copies, identical across hosts.
  • A fixture or skill test covers: a nonzero result, a success, a missing outcome, repeated copies of one result, and the same call id in two unrelated sessions.
  • Out of scope: normalizing tool_status in product code. That is a separate decision.

🤖 Generated with Claude Code

Lingua principale
JavaScript
Stelle
11
Fork
0
Merge medio
2h 25m
PR unite (30g)
311

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 hyparam/hypaware

Tutte le issue di hyparam/hypaware

Issue simili

Altre issue su JavaScript

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.