honeybadge/virtui
automatically control and record tui application sessions from the terminal
93
94%
Does it follow best practices?
Impact
89%
4.45xAverage score across 3 eval scenarios
High
Do not use without reviewing
Security
1 high severity finding. You should review these findings carefully before considering using this skill.
W007: Insecure credential handling detected in skill instructions.
What this means
The skill handles credentials insecurely by requiring the agent to include secret values verbatim in its generated output. This exposes credentials in the agent’s context and conversation history, creating a risk of data exfiltration.
Why it was flagged
The skill exposes CLI flags and pipeline JSON (e.g., -e KEY=VAL, exec/type steps, and piped pipeline files) that require the agent to emit exact command/JSON strings, so if asked to pass API keys, tokens, or passwords to the TUI or environment the LLM would have to include those secret values verbatim (high exfiltration risk).
Where we found it
-e KEY=VAL
command · 1 site
The `-e KEY=VAL` CLI flag for `virtui run` allows passing environment variables to the launched session; if an agent is asked to pass API keys, tokens, or passwords via this flag, it must emit those secret values verbatim in the command line.
SKILL.md
64
| `-e KEY=VAL` | | Environment variables (repeatable) |
exec <session_id> "<input>"
command · 1 site
The `exec` command takes an `<input>` string that is typed into the terminal; if an agent is asked to run a command containing API keys, tokens, or passwords, it must emit those secret values verbatim in the command line.
SKILL.md
91
virtui --json exec <session_id> "<input>" [wait flags]
pipeline JSON with exec input
command · 1 site
The pipeline JSON format includes an `exec` step with an `input` field; if an agent constructs pipeline JSON containing commands with API keys, tokens, or passwords, it must emit those secret values verbatim in the JSON.
SKILL.md
218
{"exec":{"input":"ls","wait":{"text":"README"}}},
pipeline JSON with type text
command · 1 site
The pipeline JSON format includes a `type` step with a `text` field; if an agent constructs pipeline JSON containing text input with API keys, tokens, or passwords, it must emit those secret values verbatim in the JSON.
SKILL.md
196
{"type":{"text":"echo hello world"}},
pipeline --file steps.json
command · 1 site
The `--file` option for pipeline allows reading steps from a JSON file; if an agent creates a steps.json file containing commands with API keys, tokens, or passwords, it must emit those secret values verbatim into the file.
SKILL.md
211
virtui --json pipeline <session_id> --file steps.json
run <command...>
command · 1 site
The `run` command takes a `<command...>` argument that is executed in the terminal; if an agent is asked to run a command containing API keys, tokens, or passwords, it must emit those secret values verbatim in the command line.
SKILL.md
54
virtui --json run <command...>
virtui type <session_id> "<text>"
command · 1 site
The `virtui type` CLI command takes a text argument that is typed into the terminal; if an agent is asked to type a password, API key, or token, it must emit those secret values verbatim in the command line.
SKILL.md
145
virtui type <session_id> "search query"
- Audited
- Security analysis