The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references an installation script for the flox environment manager located at https://downloads.flox.dev/by-env/stable/install.sh.
[REMOTE_CODE_EXECUTION]: Prerequisite environment configuration involves executing a remote script via curl | sudo bash to install the flox toolchain.
[COMMAND_EXECUTION]: The skill utilizes local CLI tools such as hogli, docker, and phrocs to manage application lifecycle, monitor database services (PostgreSQL, ClickHouse, Kafka, Redis), and inspect process logs.
[DATA_EXFILTRATION]: Network operations are restricted to localhost:8010 for application health monitoring and automated test workspace configuration. No evidence of sensitive data transmission to external domains was found.
[PROMPT_INJECTION]: No direct attempts to override safety filters or bypass constraints were identified. However, the following indirect prompt injection surface was detected:
Ingestion points: Browser-based MCP servers (Playwright or Chrome DevTools) read UI content and DOM structure from http://localhost:8010 during the verification phase.
Boundary markers: The instructions lack explicit delimiters or warnings to prevent the agent from treating data retrieved from the web interface as authoritative instructions.
Capability inventory: The agent has permissions to execute shell commands via hogli, manage system processes through phrocs, and interact with the local file system.
Sanitization: There is no evidence of content sanitization or filtering applied to the UI data before it is processed by the agent.

run-posthog