Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Team Tasks.Skip

v1.0.0

Coordinate multi-agent development pipelines using shared JSON task files. Use when dispatching work across dev team agents (code-agent, test-agent, docs-age...

0· 39·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
!
Purpose & Capability
The name/description match a JSON-based task orchestration tool and the included CLI/script implement that. However, SKILL.md hard-codes session keys for Telegram groups (agent:code-agent:telegram:group:-5189558203, etc.) and instructs the orchestrating agent to call sessions_send to those keys. A generic multi-agent task manager should not bake in specific external group session IDs; this is unexpected and disproportionate to the stated purpose.
!
Instruction Scope
Runtime instructions explicitly tell the agent to call sessions_send(sessionKey=..., message=...) to external session keys and to include 'working directory' and 'previous stage output' (local file paths and outputs) in dispatch messages. That directs the agent to collect and transmit local filesystem context and prior outputs to external endpoints. The SKILL.md also documents a deliveryContext workaround for sessions_send replies — together these give the skill broad discretion to transmit local data to third-party chat groups.
Install Mechanism
No install spec is provided (instruction-only plus a bundled Python CLI script). That reduces supply-chain risk; nothing is downloaded or extracted during install. The presence of a standalone script is expected for a CLI-style skill.
!
Credentials
The skill declares no required credentials or env vars, but the README/SKILL.md assume a default data/workspace path under /home/ubuntu/clawd/data/team-tasks (and optionally TEAM_TASKS_DIR). The larger concern is not missing env vars but that the instructions cause transmission of local paths/outputs and use hard-coded external session keys — effectively requesting the agent to send potentially sensitive local data to unrelated external chat groups without any declared credential or explicit user-specific configuration.
Persistence & Privilege
always is false and model invocation isn't disabled (normal). The skill does not request permanent platform-wide privileges. Note: autonomous invocation combined with the instruction to send messages to external session keys increases blast radius if the agent is allowed to run without a human-in-the-loop.
What to consider before installing
This tool appears to implement the documented pipeline features, but pay attention to these red flags before installing/using it: - SKILL.md contains hard-coded session keys (Telegram group IDs) and instructs sessions_send to those groups. Replace or remove those keys — do not let the agent send messages to unknown external groups. - The dispatch template tells the agent to include local working directories and previous-stage outputs in dispatched messages. That can leak file paths, code snippets, or other sensitive artifacts to external endpoints. Only dispatch sanitized summaries and avoid sending raw files or secrets. - The tool is single-node JSON backed (no file locking). If you intend to run multiple agents or concurrent dispatch, review scripts/task_manager.py for race conditions and add locking if needed. - Because the skill can be invoked autonomously, consider requiring a human confirmation step in your agent workflow before any sessions_send or network-bound dispatch occurs. Practical next steps: 1) Inspect scripts/task_manager.py fully (search for network calls, subprocess.exec, file reads of sensitive paths). Run it locally in a sandboxed/isolated environment first. 2) Remove or parameterize the hard-coded session keys and verify sessions_send targets are ones you control. Prefer configurable session keys via environment variables rather than baked-in IDs. 3) Use a dry-run mode (or instrument the script) to print messages instead of actually sending them to external sessions while you validate behavior. 4) If you plan to share agent outputs, explicitly sanitize or redact file contents and secrets before including them in dispatch messages. Given the explicit external session targets and instructions to transmit local context, treat this skill as potentially leaking data until you verify and reconfigure it.

Like a lobster shell, security has layers — review code before you run it.

latestvk97d6aqaextpy9qv2r9mwnkdes83z2nv

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments