aws / aws/amazon-q-developer-cli

Subagent `read` tool silently lacks Image mode despite being listed as available

Aperta
#3,658 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub
Lingua principale
Rust
Stelle
2k
Fork
439
Metriche di merge delle PR
Nessuna PR unita negli ultimi 30g

Descrizione

### Checks

- [x] I have searched [github.com/aws/amazon-q-developer-cli/issues](https://github.com/aws/amazon-q-developer-cli/issues?q=) and there are no duplicates of my issue
- [x] I have run `q doctor` in the affected terminal session
- [x] I have run `q restart` and replicated the issue again

### Operating system

macOS 15.7.4 (24G517)

### Expected behaviour

The read tool's Image mode should work identically in subagents and directly-invoked agents. When a subagent calls read with Image mode on an image file, it should return the image content for the model to analyze — the same behavior as a direct kiro-cli chat --agent invocation.

If Image mode is intentionally unsupported in subagents, it should be listed under "Not available" in the subagents documentation alongside web_search, grep, glob, etc., and the tool should return a clear error message rather than silently degrading.

### Actual behaviour

When a subagent attempts to read an image file using the read tool's Image mode, it fails with "no read tool with Image mode available in current environment." The subagent cannot process any image files (PNG, JPG, etc.) and falls back to treating them as text, resulting in UTF-8 encoding errors or outright refusal.

This occurs even when the subagent's agent configuration explicitly includes `read` in both `tools` and `allowedTools`, and the same agent succeeds when invoked directly via `kiro-cli chat --agent`.

The subagents documentation lists `read` as an available tool, and the built-in tools reference describes `read` as "Reads files, folders and images" — but the Image mode is silently missing in the subagent runtime. There is no error or warning indicating the capability is unavailable.

### Steps to reproduce

1. Create a custom agent with `read` in its `tools` and `allowedTools`
2. Place an image file (PNG/JPG) in the working directory
3. Run the agent directly: `kiro-cli chat --agent ` — ask it to read the image with Image mode. It succeeds and describes the image.
5. From a separate agent, invoke the same agent as a subagent via `use_subagent` and ask it to read the same image. It fails, reporting that Image mode is not available in the current environment.

### Environment

```yaml
[q-details]
version = "1.26.2"
hash = "a8687141fdf1e599719c4da99f88f338aba93537"
date = "2026-02-18T03:22:58.712286Z (11d ago)"
variant = "full"

[system-info]
os = "macOS 15.7.4 (24G517)"
chip = "Apple M4 Pro"
total-cores = 14
memory = "48.00 GB"

[environment]
os = "Mac"
shell-path = "/bin/zsh"
shell-version = "3.2.57"
terminal = "iTerm 2"

[env-vars]
SHELL = "/bin/zsh"
TERM = "xterm-256color"
```

Guida per i contributori

Apri la guida per i contributori

Direzione di ricerca

Inizia riproducendo la differenza tra `kiro-cli chat --agent` e l’invocazione dello stesso agente tramite `use_subagent`, usando Image mode dello strumento `read` con un PNG o JPG. Leggi la documentazione di subagents e il riferimento agli strumenti integrati; il lavoro è terminato quando Image mode funziona nei subagents oppure la sua indisponibilità è documentata e produce un errore chiaro.

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

Valutazione

Stack tecnologico
rust
Ambito
ai, cli
Tipo di issue
Bug
Difficoltà
4/5
Tempo stimato
3-5 giorni
Stato di attività
Ferma
Chiarezza
Abbastanza chiara
Idoneità per principianti
38/100

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.