Security audit
Security checks for vulnerabilities and agentic risk
The skill is mostly a disclosed multi-agent build workflow, but it includes live deployment, symlink swapping, and service restart instructions without clear approval or environment safeguards.
Review this before installing if you use agents near live services. The coordination parts are ordinary, but treat the deployment section as requiring explicit human approval, confirmed target environment, backups, and rollback planning before any copy, symlink, config, or restart action.
The documented actions include copying files into live deployment paths, swapping symlinks, and restarting services. These are privileged operational capabilities that can alter production code paths, cause outages, deploy incomplete/incremental builds incorrectly, or overwrite trusted files if the paths are wrong or attacker-influenced.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
Frontend: 🔨 Building
## Backend Log
<!-- Only the backend agent edits this section -->
- 2024-01-15 10:00: Started router implementation
- 2024-01-15 10:30: GET /api/v1/items endpoint complete
- 2024-01-15 11:00: BLOCKER: Need auth middleware from infra team
The skill is presented as a collaboration/orchestration protocol, but it explicitly instructs the agent to perform deployment, change live server targets, and restart services. That scope expansion is dangerous because an orchestration skill may be invoked in routine build contexts, yet it contains production-affecting actions that can modify or interrupt a live system without an explicit, separate authorization boundary.
The skill provides operational instructions that can impact availability and integrity of a live service—copying files, changing deployment targets, and restarting the application—without strong warnings, approval gates, or discussion of service/data risk. In practice, this normalizes high-impact actions inside a routine build protocol and increases the chance of accidental production changes or downtime.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
# Test each endpoint from the spec
curl -s http://localhost:8000/api/v1/{resource} | jq .
curl -X POST http://localhost:8000/api/v1/{resource} -d '{...}' | jq .
Line L108 asserts that no simultaneous-write conflicts are possible because each agent has its own log file. But L199-L204 explicitly describes a scenario where both agents edit integration.md at the same time and create a conflict, contradicting the earlier claim.
No suspicious patterns detected.