Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Audio Announcement Skills
v1.7.4Enables AI agents to announce their real-time actions via voice in multiple languages, with queued, concise, and friendly audio updates for tasks and status.
⭐ 0· 71·0 current·0 all-time
by@wililam
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
Name/description align with included scripts: bash and Python scripts implement text-to-speech using edge-tts (online) and local fallbacks. The package metadata and README are consistent with a cross-platform TTS announcement skill. Minor oddity: package.json lists 'python' and Python libraries as dependencies in an npm-like manifest, but this is a packaging/metadata quirk rather than a functional mismatch.
Instruction Scope
SKILL.md explicitly instructs operators to add enforced voice broadcasting to AGENTS.md so the agent will 'must' speak on every interaction, including a one-line spoken summary immediately before sending text replies. This is scope-expanding and privacy-relevant: it forces the agent to vocalize content (potentially sensitive) and the docs encourage modifying agent config to make that binding. Additionally, workflow-helper.sh provides wrappers that eval/exec commands (announce_with_progress uses eval on passed step_cmd strings), which can run arbitrary system commands if used — expected for a wrapper script, but potentially dangerous if misused or used with untrusted inputs.
Install Mechanism
No automated install spec in the registry entry; the SKILL.md instructs installing via ClawHub or manual git clone from a GitHub repo. All code is included in the package, so nothing will be fetched at install time by the registry itself. The recommended runtime dependency edge-tts and optional pygame/ffmpeg/mpg123 must be installed by the operator (pip/apt/powershell), which is standard for this type of skill.
Credentials
The skill does not request environment variables or credentials in metadata (none declared). The runtime uses edge-tts which performs network calls to generate speech (no credentials visible in code). It caches generated MP3s under ~/.cache/audio-announcement and temp files under /tmp or the OS temp dir; KEEP_TEMP_FILES is enabled in the Python player so audio may persist. No direct credential exfiltration is present, but network access and local caching are required for normal operation.
Persistence & Privilege
It does not set always:true or request elevated privileges. However the documentation asks operators to edit AGENTS.md to enforce speaking on each interaction — a configuration change that grants the skill persistent behavioral influence over the agent. The skill writes to user-local cache and temp paths (~/.cache, /tmp) and may leave temp mp3 files if KEEP_TEMP_FILES is true.
What to consider before installing
This skill appears to implement what it claims (local and network TTS), but consider the following before installing: 1) Privacy: the docs insist the agent 'must' speak on every interaction and even before sending text replies — that can cause sensitive information to be vocalized aloud; only enable this if you accept that risk. 2) Network & dependencies: it relies on edge-tts (network TTS) and optional players (ffmpeg/mpg123/pygame); ensure you are comfortable with outgoing network calls and install the needed packages or use the offline script. 3) Local artifacts: audio is cached in ~/.cache/audio-announcement and temp directories; review and clean if desired (KEEP_TEMP_FILES in the Python player retains files). 4) Command execution: workflow-helper.sh includes helpers that will eval/exec commands you pass in — do not use these wrappers with untrusted input. 5) Configuration change: the skill asks you to modify AGENTS.md to force behavior — review that change carefully and consider testing in a safe environment first. If any of the above is unacceptable, do not enable the forced broadcasting; you can still use the scripts manually or in offline mode after reviewing the code.Like a lobster shell, security has layers — review code before you run it.
latestvk974tnatqgq8jn6wvv13ke2h0983kx53
License
MIT-0
Free to use, modify, and redistribute. No attribution required.