The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill is designed to automatically detect and execute scripts defined in project files such as package.json or docker-compose.yml. This creates a direct path for executing arbitrary code if the project directory contains malicious configurations.
Evidence: SKILL.md Workflow step 4 and 'Environment Detection' section describe executing detected package managers and docker compose commands based on local file presence.
[PROMPT_INJECTION] (HIGH): The skill is highly susceptible to Indirect Prompt Injection (Category 8). It relies on untrusted external data (the project's files) to make decisions about command execution.
Ingestion points: package.json, docker-compose.yml, lockfiles, and process metadata retrieved via lsof and ps (referenced in scripts/check_ports.sh).
Boundary markers: None. There are no delimiters or instructions to ignore embedded commands within the processed project files.
Capability inventory: Process termination (kill -9), arbitrary shell execution via package managers, and directory navigation.
Sanitization: None. The skill does not validate the integrity or safety of the scripts it discovers before suggesting or performing execution.
[COMMAND_EXECUTION] (MEDIUM): The provided scripts/check_ports.sh script gives the agent the ability to terminate running processes using kill -9.
Evidence: The kill_port function in scripts/check_ports.sh executes hard kills on PIDs identified via lsof. While a 'same project' check is implemented, this logic relies on cwd which can be manipulated or misinterpreted.

dev-server