Agent Persona Exploration - 2026-11-12

Aperta Adatta ai principianti
  • Qualcosa non va?
#77,188 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
2/5
Tempo stimato
1-3 ore
Idoneità per principianti
72/100
Tipo di issue
Documentazione
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
github-actions, go
Ambito
ci-cd, devtools

Direzione di ricerca

Inizia con le tre raccomandazioni nell’issue: .github/aw/network.md e le indicazioni intorno alle righe 103 e 255 di .github/aw/create-agentic-workflow.md. Leggi quelle sezioni e confrontale con i risultati riportati sopra; il lavoro sarà completo quando la documentazione tratterà esempi di ricerca nel registro, il preelaborazione deterministica per i controlli strutturali e l’inserimento dei valori predefiniti nei riepiloghi dei prompt.

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

Descrizione

agent-research cookie
Persona Overview
  • Agent: agentic-workflows
  • Personas This Run: Frontend Developer, DevOps Engineer, QA Tester
  • Rotation: all 9 personas are now covered, so the rotation resets next run
  • Scenarios Tested: 4 of 6 generated (FE-5, DO-5, DO-6, QA-5); FE-6 and QA-6 were not tested
  • Invocation: one persona-evaluator run; all 4 scenarios returned recommendations
  • Average Quality Score: 4.5/5.0
  • Run: §43248107250
Key Findings
  • Triggers were correct in all four recommendations: pull_request with path filters for PR reviews, and schedule plus workflow_dispatch for the scheduled check. Trigger appropriateness averaged 5.0.
  • All four used read-only permissions, no edit tool, pull_request rather than pull_request_target, and COMMENT-level reviews that do not block PRs.
  • The weakest dimensions were completeness (4.0) and prompt clarity (4.25). Defaults the agent chose (exemption matching, the 48-hour cron rule, Go subtest rules) were written in the notes but often missing from the prompt summary.
  • DO-6 was the only over-broad network suggestion. It hardcoded four registries, although the containers ecosystem already covers GHCR and Docker Hub.
  • No scenario averaged below 4.4.
Top Patterns
  1. Triggers: pull_request with paths: filters (3 of 4); schedule plus workflow_dispatch (1 of 4).
  2. Tools: github gh-proxy and bash in all four. Shell commands handled file listing, test scanning, and CLI lookups.
  3. Safe outputs: review comments (FE-5, QA-5), a rolling add-comment with hide-older-comments (DO-6), and create-issue with close-older-issues (DO-5). noop appeared in all four.
View Scenario Scores
Scenario Task Trigger Tools Security Prompt Completeness Avg
FE-5 Storybook story gap on component PRs 5 5 5 4 4 4.6
DO-5 Scheduled workflows silent for 48h 5 5 5 4 4 4.6
DO-6 Unpinned Dockerfile base images 5 4 4 5 4 4.4
QA-5 Test functions with no assertions 5 4 5 4 4 4.4
View High Quality Responses
  • FE-5 (4.6): Inline COMMENT-level review per component missing a story, honors the exemption list, caps at max: 25, and calls noop when nothing is missing. Read-only, no network.
  • DO-5 (4.6): Treats disabled workflow states as out of scope and marks crons longer than 48 hours as not evaluated rather than stale. close-older-issues with a key dedupes the tracking issue.
View Areas for Improvement
  • DO-6: Hardcodes quay.io, gcr.io, mcr.microsoft.com, and public.ecr.aws. The same response says to limit hosts to registries the repo names. It also assumes crane or docker buildx is available without checking, and does not say whether to pin the index digest or a per-platform digest.
  • QA-5: Detects test functions with grep/awk, which is brittle for Go and TypeScript. The Go subtest and t.Skip-only rules appear only in the notes.
  • FE-5, DO-5, QA-5: Exemption matching, the story file pattern, and the 48-hour cron rule are labelled defaults in the notes but missing from the prompt summaries.
Recommendations
  1. .github/aw/network.md: Add a worked example for registry-backed lookups. Prefer ecosystem identifiers such as containers, and add FQDNs only for registries that the repository files actually name.
  2. .github/aw/create-agentic-workflow.md: For structural checks (test-function detection, digest resolution), recommend deterministic steps: that write to /tmp/gh-aw/ so the agent only formats results. The guidance at line 255 covers this only for large-data preprocessing.
  3. .github/aw/create-agentic-workflow.md: Line 103 already asks for labelled Default: values. Require that they also appear in the prompt summary, not only in the notes.

References:

Generated by 🎭 Agent Persona Explorer · claude · routed: haiku55 xhigh · ~6 AIC · ⌖ ~2 AIC · ⊞ 7.6K · ◷

Lingua principale
Go
Stelle
5.6k
Fork
625
Merge medio
5h 39m
PR unite (30g)
796

Preparare l'ambiente

Apri in Codespaces

Avvia il container di sviluppo del progetto nel browser, con il tuo account GitHub.

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 github/gh-aw

Tutte le issue di github/gh-aw

Issue simili

Altre issue su Go

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.