The Agent Skills Directory

[COMMAND_EXECUTION]: The skill enables the agent to execute the node-devtools-cli utility, which provides significant control over Node.js processes on the system.
Evidence: The allowed-tools section in SKILL.md grants access to all node-devtools-cli subcommands via the Bash tool.
Evidence: The utility can signal processes (SIGUSR1), connect via WebSockets, and evaluate JavaScript expressions through commands like node-devtools-cli debug add-watch --expression and node-devtools-cli debug put-logpoint --log-expression.
[EXTERNAL_DOWNLOADS]: The skill documentation refers to an external package that must be installed on the host system.
Evidence: SKILL.md instructs users to run npm install -g browser-devtools-mcp to install the necessary CLI tools.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection by processing data from debugged applications.
Ingestion points: The agent retrieves potentially untrusted data through debug get-logs and debug get-probe-snapshots in references/debug.md.
Boundary markers: The documentation does not specify the use of delimiters or instructions to ignore embedded commands within the captured logs or snapshots.
Capability inventory: The agent has the ability to execute shell commands and modify process state through the CLI tools defined in SKILL.md.
Sanitization: There is no mention of sanitization or filtering of the content retrieved from the Node.js process before it is presented to the agent.

node-devtools-cli